Digital Marketing

RETAIL AUTOMOTIVE: An Industry Unprepared

| admin

The second in a three part series explaining how the retail Automotive industry appears unaware of the paradigm shift in AI search summaries, citation and search behaviour – and are not implementing change to accomodate a shift they are mostly unaware of.

Why your dealer website provider isn’t ready for AI search, and why that’s your problem too

Every dealer website says roughly the same thing. Decades of trading. Manufacturer approved. Trusted by thousands of customers. None of that means anything to an AI agent unless it can be corroborated.

This is the part of AI search that automotive retail hasn’t caught up with yet. ChatGPT, Google’s AI Overviews and Perplexity don’t take a dealer’s word for its own expertise, trust or authority. They cross-reference it. A claim only counts if it’s backed by structured data, consistent entity information and verifiable signals scattered across the dealer’s own site and the wider web. Say it without the backing, and an AI model simply won’t repeat it, or worse, just won’t mention the dealer at all.

Corroboration, not copywriting

Most dealers still think of trust signals as a writing problem: get the tone right, mention the years in business, add a testimonials page. AI agents work differently. They check whether a claim is structurally supported, not whether it reads well.

A dealer stating compliance, regulatory or association membership needs that claim reflected consistently across the FCA and Companies House records, the Google Business Profile, manufacturer directories and the dealer’s own site, all pointing to the same verifiable entity. A dealer claiming forty years of trading needs that history to show up somewhere an AI model can check it, not just as a line on the About page. This is closer to infrastructure than marketing, and it sits squarely in technical SEO territory, which is exactly where most dealer sites are weakest. Critically, making claims that cannot be corroborated by an AI Agent will dramatically reduce the likelihood of an AI mention or citation.

Where dealer website providers are behind

Dealer platform providers built their systems for a different web. Fast stock feeds, finance calculators, lead capture forms. That’s what dealers have been sold, and it’s what most providers still optimise for.

Structured data on these platforms is typically limited to basic vehicle schema, enough to get a car listing showing correctly in a regular search result. Beyond that, the gaps are consistent across the sector: no proper Organization schema tied to a verifiable entity, no Person schema for staff or specialists, Review and AggregateRating markup either missing or poorly implemented, LocalBusiness data that’s inconsistent across branch pages and Google Business Page, and no FAQPage schema answering the actual questions buyers now put to ChatGPT rather than search – or are intercepted and answered in an AI Overview generated by Google and sitting above the results page. Never mind ensuring content is not generic and is so-called non-commodity. AI Agents tend to cite based on the content of the top 30% of a web page, particularly the first 10%, also if the page fans out with follow up questions. Then duplicated in the machine-readable code. These concepts are virtually non-existent in automotive.

A platform built to serve hundreds of dealers from one template cannot produce dealer-specific authority, because authority is not a template feature. It comes from a dealer’s own history, staff and reputation, and a generic site simply has nowhere to put that.

The signals AI agents are actually checking

The OPTIMUM Seven Dimension AI Summary and Citation Audit exists precisely because these signals need to be checked individually, not assumed. The dimensions covering technical readiness and schema implementation look at entity consistency across every external reference to the dealer, citation density on trusted third-party sources, staff and authorship credibility, review authenticity and volume, robots.txt and llms.txt configuration that isn’t accidentally blocking AI crawlers, and structured data depth on every page, not just the homepage.

Most dealer sites fail several of these dimensions without anyone noticing, because the site still looks fine to a human visitor. The problem is invisible until it’s tested against how an AI agent actually reads the page.

This is a bigger job than the industry has clocked

The honest assessment is that automotive retail is not ready for this. Fixing vehicle schema is an afternoon’s work for a competent developer. Building genuine entity consistency, credible authorship signals and page-level structured data across an entire dealer site, and keeping it that way as stock, staff and locations change, is an ongoing technical and editorial commitment. Most dealer groups have neither budgeted for it nor assigned anyone to own it, and most platform providers are treating it as a features list item rather than the structural rebuild it actually is.

That gap matters because the timeline is short. The wider thesis behind this shift, ‘Search Doomsday’, points to Q3 2027 as the point where AI answers overtake traditional click-through search as the default buyer journey. That is not a distant horizon. It’s roughly a year away, and the work required here is not the kind that gets done in a sprint. It will affect any industry where informational search queries are the first part of a prospect’s journey.

What the dealer actually has to do

None of this can be outsourced entirely to a platform provider, however good the provider is. The dealer is the entity being corroborated, so the dealer has to own the consistency of that entity everywhere it appears. That means auditing what currently exists, fixing trust and schema gaps page by page, and treating technical SEO as a live discipline rather than a one-off build.

An OPTIMUM Citation Gap Analysis is the starting point for any dealer wanting to know where they currently stand, rather than assuming their website provider has this covered. Most haven’t.

State of the Art Digital works with automotive retailers on the OPTIMUM Seven Dimension AI Summary and Citation Audit. Contact steve@stevecoulter.co.uk or +44 (0)7407 038877.

AI SEARCH REPORT: The Fan Out Effect

| admin

The Fan Out Effect and AI Citations

Executive summary

AI search is changing how brands are discovered and referenced. The main lesson from recent research into query fan out is that citation in AI answers depends less on broad authority and more on whether a page is retrieved early, matches the query closely, and is structured in a way the model can use.

For State Of The Art Digital, the practical takeaway is clear. Content strategy now needs to be built for citation as well as ranking. That means sharper page intent, better heading alignment, and a stronger focus on direct answers rather than broad topic coverage.

Report overview

Recent industry research into query fan out examines how AI systems move from a user prompt to the sources they cite. The analysis draws on a large set of queries and retrieved pages, giving a useful picture of how citation decisions are made in practice.

The findings show that AI search does not work like a simple keyword ranking system. Instead, the model expands a prompt into related sub-queries, retrieves a broad set of pages, and then narrows down to the sources that best fit the answer.

What the data suggests

The strongest signal appears to be retrieval position. Pages that surface near the top of the retrieval set are much more likely to be cited than pages that appear lower down. In other words, if a page is not visible early in the retrieval process, it is unlikely to feature in the final response.

Heading relevance also matters a great deal. Pages whose headings closely match the user’s query are cited more often than pages with weaker or more generic section titles. This suggests that clear, question led structure is a practical advantage, not just a stylistic preference.

The research also indicates that traditional authority signals do not carry the same weight here as they do in standard SEO. Domain authority and backlinks may still help overall visibility, but they do not seem to be the deciding factor when AI systems choose which page to cite.

Why this matters

This changes how content teams should think about optimisation. Long, all-purpose pages are not automatically better, especially if they dilute the relevance of the page to one specific question. A tighter page that answers one intent directly may have a better chance of being surfaced and cited.

It also means that content quality alone is not enough. A well written page still needs to be easy for the model to interpret, with headings, structure, and topical focus that closely mirror the user’s likely prompt.

Practical implications

For brands that want to improve AI visibility, the first priority should be page alignment. Each important page should be built around one clear intent, with headings that reflect the way people actually ask the question.

The second priority is structure. Short, well ordered sections make it easier for AI systems to identify useful passages and extract them into an answer. This is especially important for service pages, FAQ sections, and comparison content.

The third priority is clarity over breadth. Rather than trying to cover everything on one page, it is often better to create focused pages that answer a single task or question properly.

Recommendations for clients

  • Build pages around one primary search intent.
  • Use headings that closely match real user questions.
  • Keep sections concise and logically ordered.
  • Refresh important pages regularly so they stay current.
  • Support core pages with internal links and related content.
  • Treat AI citations as a visibility goal alongside organic rankings.

Conclusion

The key message from the research is that AI search rewards precision. Brands are more likely to be cited when their content is easy to retrieve, easy to interpret, and clearly matched to the prompt being answered.

For State Of The Art Digital, this is an opportunity to position clients for the next phase of search visibility. The winning approach is no longer just to rank, but to become the clearest source a model can confidently quote.

Please contact me if you would like your business website structure and navigation adapted for your industry to suit early AI bot retrieval.

AI SEARCH: Fake Plastic Models

| admin

Fake Reviews Aren’t for People Anymore. They’re for the Models.

Reddit is being flooded with AI-generated posts and reviews. According to a recent MediaPost piece, brands are planting them there because ChatGPT and Google’s AI treat Reddit as a trusted source of real human opinion. The target isn’t a reader. It’s an algorithm looking for something to cite.

If that sounds familiar, it should. It’s the SEO playbook from the 2000s, one layer up.

Back then, ranking on Google was something you gamed: keyword stuffing, invisible text, link farms, thousands of dodgy backlinks. It worked, for a while. Then Google shipped Panda and Penguin, and most of those tactics stopped working overnight. Sites built on tricks collapsed. Sites built on genuinely useful content held steady.

We’re watching the same cycle repeat, with forums instead of backlinks. The models now read Reddit as ground truth, so that’s where the fake signal gets planted. And the platforms are already fighting back. Reddit is blocking 23 million spam views a day, removing close to 2 million fake votes daily, and catching around 25,000 spammy posts every day. The enforcement window that took Google years is now measured in seconds.

The reason this is happening is simple: search has moved into the answer box. Zero-click queries rose from 56% to 69% in under a year, and news sites lost 600 million monthly visits in the process. Being findable now means being cited, and brands are scrambling to work out how.

The lesson from the SEO era still applies: astroturfing is the wrong answer to the right question. It works until the platform catches it, and platforms are catching it in seconds now. You’re building your visibility on someone else’s moderation policy.

The better question is the one Panda and Penguin eventually forced everyone to ask: why would an AI cite you at all? Usually because the information about your product is clear, structured and genuinely useful, not because you gamed a forum.

Keyword stuffing didn’t survive. Fake reviews won’t either. The brands that win the citation race will be the ones the models actually understand, not the ones spamming Reddit.

AI Search Doomsday: Don’t Look Up!

| admin

Ever seen the film Don’t Look Up? A black comedy drama which portrays a group of astronomers (including; Leo DiCaprio & Jennifer Lawrence)  who accidentally notice a distant asteroid is on a collision course with Earth that might be diverted – but fail to get their warning across to a sceptical world and leaders. Inevitably civilisation ends. There’s a lot of running around near the end.

AI Search Doomsday Q3 2027 won’t end the world, but my detailed and peer reviewed research suggests it could end or do serious damage SME and MME businesses who are not already working on increasing the likelihood of their web pages’ AI Summary inclusion and citation or appearing on AI App outputs. Few are being assisted by the website provider industry – the reluctant gatekeepers of change

The people who I have spent time with over the last three months are truly the outliers and thinkers in industry. They have noticed an unprecedented and difficult to interpret drop in Click Through Rates from vital search phrases and/or recognise the threat AI Search poses to their industry. We have a remedies underway. Where is your business on this paradigm shift?

My Linked-In post today.

In late 2024 I started digging into a question most business owners haven’t asked yet:

What happens to search when AI stops sending clicks to websites and just answers the question itself.

The authoritative and published data I’ve been tracking since is stark. Zero-click searches have gone from 56% to 69% of all Google queries in a year. When an AI Overview appears, that jumps to 83%. Organic click-through rates on those queries are down 61%. HubSpot, one of the best SEO operations in the world, lost 70 to 80% of its organic traffic in under 12 months, while its rankings barely moved.

That’s the part people are starting to notice. What I’ve spent the last year and a half researching is what it actually means for a business, and I’ve written it up across three reports which are all now published.

– Your Business Is Becoming Invisible sets out the diagnosis. Ranking and being cited by AI are now two separate problems. Across 20 UK automotive and estate agency audits I ran, the average AI readiness score was 5.5 out of 10, and not one of those businesses knew it.

– Search Doomsday Is Q3 2027 puts a date on it. AI is currently intercepting around 17% of the clicks that would once have gone to a website. Every data source I’ve reviewed converges on the same figure: 40% by Q3 2027. That is the point where the question changes from “do we rank?” to “do we get cited?”

– The Chaff Effect is the sharpest and damning finding of the three. Clicks are not disappearing evenly. AI resolves informational, early-stage searches 74% of the time, but only 31% of transactional ones. Businesses are losing the awareness-stage traffic that used to feed everything else, while the pool that is left quietly runs dry behind them.

Next week I’m releasing a benchmark that lets any SME or MME leader, and the marketers working for them, see exactly where they stand against this shift, and what to do about it before the inflection point arrives.

If your enquiries have been sliding while your rankings look fine, this is why.

More next week.

—

Feel free to contact me if you are the director or employee in the business pushing to ensure yours is included in AI Summaries, Citations and on AI Apps.

AI SEARCH: Why Small Brands Still Have a Chance

| admin

For brands also read businesses. Your business.

Apple, Reddit and HubSpot are not scrapping for a place in AI answers, and there is one reason for this: training data.

AI engines will cite and quote a brand under two conditions:

  1. The model already knows the brand (training data)
  2. The model can find the answer it needs on the brand’s website (retrieval)

Large brands have already secured their position in the AI citation game. They dominate answer engine optimisation for free, more or less, because these models were trained on data in which those brands are already deeply embedded.

Sixty per cent of ChatGPT answers never touch the live web at all, which tells you how much weight training data carries.

For a smaller brand, competing on training data is a losing battle. The real opportunity lies in the second path: retrieval.

When an AI engine cannot find the answer in its training data, or needs something current, it runs a web search. This is where a smaller brand gets its chance to be cited and quoted. The route to that chance has not changed: SEO. But it is a specific flavour of SEO that matters here.

Retrieval depends on technical foundations:

Technical SEO. If a crawler cannot access, render or parse your site cleanly, none of your content is retrievable, no matter how good it is. Fast load times, clean site architecture, proper canonicalisation and a crawlable structure are the entry ticket, not an afterthought.

Structured pages. Content built around one clear question per page, with a direct answer near the top, gets pulled into AI responses far more easily than content buried in long, unfocused articles. Structure your pages the way you would structure an answer if someone asked you directly.

Comprehensive schema markup. Schema tells AI engines exactly what your content means, not just what it says. FAQ schema, Article schema, Organisation schema and Product schema, used properly and thoroughly across the site, give models the clearest possible signal that your page holds the answer they are looking for.

llms.txt. This is the newest lever available. An llms.txt file gives AI models a direct, structured route to your most important content, much as a sitemap does for traditional search engines. It will not guarantee a citation, but it removes friction and makes your best answers easier to find and trust.

There is a genuine glimmer of hope here. Retrieval can beat fame. A smaller brand with a more retrievable answer, backed by solid technical SEO, well-structured pages and thorough schema, will win the citation over a bigger brand that only earns a passing mention.

The golden rule for winning at retrieval is simple:

Own your niche completely. Every topic. Every subtopic. Every question a customer might ask. The instant an AI model spots a gap in your coverage, it moves on and cites whoever has filled that gap instead.

AI Search: The Six Vital Content Traits For Citation

| admin

There is a question that more business owners are starting to ask, and it matters more than most of them realise.

When a potential customer opens ChatGPT, Claude or Perplexity and asks which local solicitor handles commercial leases, which car dealer in Norwich carries approved used BMWs, or which estate agent consistently sells in their road, whose name comes up?

In most categories, the same handful of businesses get cited again and again. Everyone else is invisible.

This is not random. AI engines do not pick names out of the air. They pull from pages that are structured in a very specific way. If your website is not built that way, it will not be quoted, regardless of how long you have been in business or how good your Google ranking is.

The question is: what does a citable page actually look like?

Having studied in detail how AI search works across multiple platforms, six traits appear on every page that earns consistent citations. Miss one and your chances drop. Miss three and you disappear entirely.

Trait one: The heading is the real question

Not a clever marketing line. Not a vague label. The actual question your customer would type.

If someone asks an AI which estate agents in Worthing sell the most family homes, your page heading should reflect exactly that. “Our services” does not get you cited. “Which estate agents sell the most family homes in Worthing?” might.

AI engines match queries to headings before they read a single word of your copy. If the heading does not fit the question, the page is skipped.

Trait two: The first sentence answers the heading directly

No warm-up. No scene-setting. No “it is a great question” preamble.

The answer goes in the first sentence. Two sentences at most. Then you can expand.

Most business websites do the opposite. They spend three paragraphs building context before they say anything useful. By then, the AI has moved on to a competitor who answered in line one.

Trait three: One specific fact that only you own

This is the one that most businesses overlook, and it is probably the most important.

Generic claims do not get pulled. “We have years of experience” earns nothing. “We have sold 47 properties within half a mile of the seafront in the last 18 months, with an average of 11 days to offer” earns citations.

The fact does not need to be dramatic. It needs to be specific, true, and yours alone. A price. A ratio. A timeline. A count. A measured outcome from your own business.

AI engines are looking for something they can lift and use without having to verify it against ten other sources. A first-party number is exactly that. A generic claim is not.

Trait four: A real person behind the page

A named author. A photograph. A bio that says something specific about their experience. Ideally a small piece of markup behind the scenes that tells AI platforms this page was written by an identifiable human being, not generated by a machine.

Claude and Perplexity both weight this signal. A page with a named author who has a verifiable background gets more trust than an identical page attributed to a faceless brand.

This is straightforward to fix if you have not done it. Add your name to the pages that matter. Write three sentences about your actual background. Make it specific. “20 years in automotive retail, including 12 years managing franchised BMW and Audi sites in the south east” is useful. “Passionate about cars” is not.

Trait five: A structure the AI can scan

Short paragraphs. Clear subheadings. The occasional list where it genuinely helps. White space.

A 900-word page with seven tight sections consistently outperforms a 1,400-word page with three sections and one dense block of copy beneath each, because the engine can locate the relevant passage quickly rather than working through a wall of text.

This is not about dumbing down your writing. It is about making it easy to extract. The underlying thinking can be as sharp as you like. The structure needs to let the engine find the answer without having to dig.

Trait six: No filler

Every filler phrase weakens your entire page, not just the sentence it appears in.

“In today’s competitive landscape.” “Navigating the world of.” “We are committed to delivering excellence.” “At the heart of everything we do.”

AI platforms have learned to classify these phrases as low-signal text. When they appear, the surrounding paragraphs lose credibility. The page reads as generic content, produced to fill space rather than answer a question.

Cut them. Every single one.

What this means in practice

Run through your three most important pages, the ones a customer lands on when they are close to making a decision.

Read the first sentence under each subheading. Does it answer the heading directly? If not, rewrite it so it does. Then add one specific first-party fact to each section. Then check whether a named author with a real bio appears on the page.

That is an afternoon’s work. Not a redesign. Not a new content strategy. Just making what content you already have citable.

The businesses that show up when a potential customer asks an AI for a recommendation are not necessarily the biggest, the oldest, or the best-ranked on Google. They are the ones whose pages are easy to quote.

At the moment, most of your competitors have not made that adjustment. That gap will not stay open for long.

Steve Coulter is a GEO and AI search consultant at State of the Art Digital. He works with automotive retailers, estate agents and professional services firms on AI visibility strategy.

About AI Citation Gap Analysis

| admin

One of the biggest misconceptions in the GEO market is that AI citations can be treated like traditional search rankings.

They can’t.

AI systems do not operate like search engines. You cannot buy ‘AI Citations for £199 per month’ – Citations are influenced by a complex mix of trust, authority, entity recognition, content quality, technical accessibility and third-party validation. No credible provider can guarantee when, where or how often a business will be cited.

This understanding sits at the heart of OPTIMUM AI Citation Gap Analysis.

Rather than chasing citation guarantees, OPTIMUM identifies the gaps across a business’s digital footprint that may affect its ability to be retrieved, grounded and cited by AI systems.

The project has been in development significantly longer than many GEO products currently entering the market, with research and methodology established well before AI visibility became a mainstream marketing trend.

The goal has never been to sell hype.

It has always been to provide businesses with a defensible, evidence-based assessment of the factors that influence AI discoverability and authority.

In a market increasingly crowded with promises, measuring reality matters more than ever.

The OPTIMUM Ecosystem offers a cost effective solution to this problem, it analyses your current AI citation position and supplies an action plan to move your business into the citable sources pool.

AI Search: Welcome Validation From The CEO of Google

| admin

Google Zero? I Said So Months Ago. Now Google CEO Has Confirmed It.

Sundar Pichai has now said publicly what anyone paying close attention already knew. Google Zero is not a conspiracy theory. It is a direction of travel that the CEO of Google has validated.

I have been writing about this for months. Not as speculation. As a reading of observable data that pointed one way and kept pointing the same way regardless of how many times the industry and experts tried to reassure otherwise.

AI Overviews now trigger on close to half of all queries, with zero-click rates reaching 80 to 93 percent in some search modes. Publishers are reporting traffic losses of 26 to 55 percent. This is not a future risk. It is a present condition, and it has been for some time.

What Pichai’s interview adds is not new information. It is confirmation from the top. He acknowledged that AI Overviews can be more opinionated than they should be. He admitted the product is still evolving. He leaned on 25 years of user satisfaction data as evidence that Google will course-correct.

With AI search summaries becoming increasingly more accountable, Pichai’s concession, alongside Google’s expansion of Preferred Sources across AI search, points toward a model where verified, trusted, authoritative sources receive preferential citation treatment.

That is the architecture GEO strategy is built around. Not ranking. Citation. Not position one. Source selection.

If your strategy still treats ranking as the end goal, this is your confirmation that the end goal has moved. The work now is becoming the source AI chooses to reference, not just the page that earns a click.

It’s good to have validation, even if somewhat reluctant and late coming.

—

* Blbliography for this article is available should you wish to conduct further research.

* For anyone unfamiliar I’m drafting a description of ‘Google Preferred Sources’ and what that means.

AI SEO: Video Is the Untapped AI Citation Asset Most Local Businesses Are Ignoring

| admin

Punch Above Your Weight With This Two-Presence Video Strategy

Most car dealers and estate agents have been producing video for years. Walk-around stock videos, branch and forecourt tours, meet-the-team clips, market update commentaries. The content exists. The problem is almost none of it is configured to be read by an AI.

That distinction matters enormously right now.

AI search platforms – ChatGPT, Perplexity, Gemini, Google AI Overviews – do not watch video. They read the text surrounding it. They parse the title, the description, the transcript, the structured data markup, and the page context the video sits within. If those elements are absent, incomplete, or inconsistent, the video is invisible to every AI system regardless of its production quality or view count.

This is the gap that presents an immediate competitive opportunity for any local business willing to spend a few hours getting the fundamentals right.

Why YouTube Dominates AI Citation, and How That Helps You

YouTube is currently the single most-cited domain across all major AI platforms. Research from early 2026 shows it appears in roughly 16 per cent of LLM-generated answers, well ahead of any other source. This is not because AI systems are watching the videos. It is because YouTube enforces consistent metadata, generates automatic transcripts, and provides structured, machine-readable content at scale.

The implication for local businesses is significant. A YouTube channel is not just a video hosting platform. Configured correctly, it is a citation asset feeding into every major AI system simultaneously. Your video description, your chapter timestamps, your pinned comment and your auto-generated or manually uploaded transcript are all indexable text that AI crawlers can extract and attribute.

The key is understanding that the same optimisation logic applies to your own website. YouTube gives you citation reach. Your own site gives you citation authority and SEO credit. The winning strategy uses both, with a deliberate canonical structure connecting them.

The Canonical Problem Nobody Is Solving

The most common video mistake local businesses make is treating YouTube and their own website as two separate, unconnected things. A video goes on YouTube. Someone embeds it on a web page. Neither has proper metadata. Neither has a transcript. There is no structured data. The two versions compete with each other in search, and neither builds authority.

The correct approach is to establish a canonical video page on your own website and treat everything else as supporting distribution. Each video gets a dedicated page with a clear, keyword-informed title, a substantive description written in full sentences, a complete transcript published as readable text, VideoObject schema implemented in JSON-LD (Javascript Object Notation for Linked Data), and the YouTube embed as the playback mechanism.

The VideoObject schema uses the canonical page URL as its @id, which signals to search engines and AI crawlers that your site owns this content. The YouTube channel amplifies reach and feeds AI citation platforms. Your site gets the SEO equity.

This dual-presence model is the structural backbone of effective video GEO for local businesses.

What AI Systems Are Actually Reading

Understanding what an AI system extracts from a video page clarifies exactly what you need to produce. When ChatGPT, Perplexity or Google’s AI Mode retrieves a page containing a video, it is reading several distinct text layers.

The first is the page title and H1 heading. These should answer a specific, naturally phrased question. Not “Ford Focus Walkround July” but “What specification is a Ford Focus 1.0 EcoBoost? A full walk-around and honest assessment.”

The second is the video description. On YouTube this needs to be at least 200 words and should front-load the most important information. AI systems give disproportionate weight to the first third of any page’s content. The same description, or a fuller version of it, should appear on your canonical web page.

The third layer is the transcript. This is the most underused asset in local business video SEO. A 90-second walk-around video contains 150 to 200 words of spoken content. Published as visible text on the page, that content becomes indexable, citable, and attributable to your business. For a market commentary video from an estate agent, the spoken words represent genuine information gain – the kind of factual, expert content that AI systems prefer to cite.

The fourth layer is structured data. VideoObject schema implemented in JSON-LD tells AI crawlers and search engines precisely what the video contains, when it was published, how long it is, who produced it, and what page should be treated as the canonical source. Without it, AI systems are guessing at context. With it, they have a machine-readable brief. Fabulous entity and topical, semantic signals for AI citation uplift.

The Local Business Advantage

Large national brands have video teams, SEO departments and agency relationships. A used car dealer in West Sussex or a three-branch estate agent in Essex is not competing with them directly. What local businesses have is hyper-specific local expertise and genuine informational authority in a narrow geography.

An estate agent producing a weekly two-minute video on what is happening in their local property market – pricing, stock levels, buyer activity – and publishing it with a proper transcript, VideoObject schema, and a canonical page is building exactly the kind of factual, locally specific, expert-attributed content that AI systems prioritise when answering questions like “What is the housing market like in Worthing right now?” On the canonical URL page add in extra questions and answer such as; “What are the best local Secondary Schools?” and “Where are the best beaches?”

That is an answerable query. The business that has published consistent, well-structured local content over six months will own the AI citation for it. The business that has uploaded unoptimised clips to YouTube or not at all and done nothing else will not.

The gap between those two outcomes is not one of budget or resource. It is one of consistent process.

AI Search Summaries: How Smaller Brands Are Competing

| admin

How To Get Your Brand Into AI Summaries: A Practical ‘EEAT’ Playbook

There is a version of this article that opens with a statistic about zero-click search rates, references a McKinsey report, and tells you that AI is disrupting the landscape.

You will not be reading that version.

Here is the thing that actually matters: AI search systems, whether Google’s AI Overviews, Perplexity, ChatGPT Search, or any of the others gaining users at pace, are not random. They are not black boxes that reward whoever shouts loudest. They have a logic, and that logic is remarkably close to something Google has been telling marketers for years: demonstrate Experience, Expertise, Authoritativeness, and Trustworthiness. (EEAT). The principles did not change. The stakes did.

When an AI system constructs a summary answer to a user’s question, it is making a series of editorial judgements. Which sources understand this topic? Which ones can be trusted? Which ones have said something specific and citable rather than something vague and generic? Your job, as a brand or business, is to make those judgements easy. Here is how.

Start With Positioning, Not Content

The single most common mistake brands make when trying to appear in AI-generated answers is trying to appear in too many of them. They produce content that covers broad territory. They write ultimate guides to entire industries. They want to rank for everything and end up trusted for nothing.

AI systems are not impressed by breadth. They are looking for signal, and signal requires specificity.

Consider the difference between a software company that describes itself as an all-in-one video platform and one that positions itself as the best tool for podcast editing. The first company is competing with every video tool on the internet. The second has a defined audience and a defined set of questions it can answer better than anyone else. When an AI system is asked what is the best tool for editing podcasts, the second company appears in that answer. The first probably does not.

This is not a content decision. It is a positioning decision. Narrow your claim. Own a space. The content follows from that; it does not create it.

Make Content That Is Actually Useful

Helpful content has become such an overused phrase in SEO that it has nearly lost its meaning. So here is what it actually means in the context of AI citation.

AI systems have seen every version of the generic blog post, the thin listicle, and the padding-heavy answer page. They have also seen genuinely useful writing: the forum post that actually solved someone’s problem, the how-to page that answered the tricky edge case, the comparison article that laid out real trade-offs rather than pretending every option was great in its own way.

The content that gets cited is the content that answers real questions with real specificity. Not how do I use this tool, but why does the audio desync when I import from a particular file format and how do I fix it. Not what is content marketing, but what content formats actually drive enquiries for a small professional services firm with a long sales cycle.

Your content plan should be built from questions your customers actually ask, not from keyword volume alone. Talk to your sales team. Read your support tickets. Go through your reviews. The questions are already there. Answer them with enough depth and clarity that someone with the problem right now would genuinely find it useful.

Build Pages AI Can Point To

There is a structural side to this that is often overlooked. AI systems do not just need good content; they need content in forms they can extract, attribute, and cite.

Three content types consistently perform well as AI citation targets.

Comparison pages. Comparison questions are among the most common queries AI systems receive. If you have a well-structured, honest comparison page covering your product against alternatives, you have created something AI systems can use to answer a question asked thousands of times a day. The key word is honest. Comparison pages that declare the author’s product best in every category are not useful. Pages that acknowledge genuine trade-offs are.

How-to content. Step-by-step explanations with clear sequencing and concrete actions are easier for AI systems to cite and summarise than opinion pieces or narrative articles. This does not mean how-to content cannot have a point of view; it means it should also be practical and functional.

Use-case content. Pages that describe specific applications of your product or service in specific situations give AI systems something to work with when a user’s query is about a context rather than a category. How a small accountancy firm uses project management software to handle client onboarding is more citable than a generic features page.

All of this works considerably better when supported by proper structured data. JSON-LD schema is not optional decoration. It is the vocabulary that tells AI systems what your content is, who it is about, and what it claims. If your site lacks structured data, you are asking AI systems to guess. Some will; many will not, when a better-structured competitor exists.

Use Real Visuals

AI systems with visual capabilities can process and reference visual content. More immediately, the people who train, evaluate, and ultimately trust AI systems use visuals as a quality signal.

Real product screenshots, genuine interface recordings, actual before-and-after examples, and video walkthroughs all contribute to perceived authenticity. They also make your content more useful, which loops back to the citation question. Content that helps people understand something is more likely to be cited than content that merely claims something.

There is also a simpler point here. Brands that use generic stock imagery look like every other brand. AI systems have encountered the same stock photo of a handshake or a lightbulb across thousands of websites. Real product visuals, screenshots from actual use, and genuine demonstrations stand apart. They signal that this content is about a real thing, produced by people who have actually used it.

Expand Your Presence Beyond Your Own Site

Your website is one signal. AI systems are reading many others.

Third-party mentions, reviews, press coverage, and independent creator content all contribute to the trust picture an AI system builds around a brand. When Perplexity or ChatGPT Search decides whether to include your brand in a response, it is not only reading your website. It is reading what others have written about you, in contexts you did not control and cannot directly edit.

This means PR is not separate from your visibility strategy; it is part of it. Getting covered in trade publications, being reviewed on independent platforms, appearing in podcast episodes, and being mentioned in the forums where your customers actually spend time all contribute to your perceived trustworthiness in ways AI systems can detect and weigh.

The practical implication is straightforward. Treat off-site presence as a deliberate programme rather than a nice-to-have. Identify the publications, communities, review platforms, and creators your target audience already trusts. Build genuine relationships with them. Create things worth mentioning. Earn the references rather than manufacturing them.

The Underlying Logic

Everything above serves a single purpose: making it easy for AI systems to understand what you do, trust what you say, and cite you as the source of a useful answer.

The businesses doing well in AI-mediated search right now are not necessarily the biggest ones. They are the ones already doing the work that EEAT has always demanded: positioning clearly, creating content with genuine depth, building structural credibility, and maintaining a consistent presence across the sources their audience trusts.

When AI systems can understand your positioning, trust your content, and point to specific pages you have built, the playing field levels considerably. You are not competing on budget. You are competing on clarity, depth, and genuine usefulness.

Those are things any business can build. Most simply have not started yet.

I offer an AI Risk Intelligence Briefing and Retained Advisory service to ensure your brand or business is making the most of the early-mover opportunity from AI Summary inclusion and citation. Please DM for more information and understand what this early adoption advantage is.

Below I’ve summarised my article for partner, board or C-Suite presentations.

Simple Paragraph Summary Bullet Points:

Start With Positioning, Not Content

  • Trying to rank for everything means being trusted for nothing
  • AI systems reward focused, specific positioning over broad claims
  • Decide what question you want to answer better than anyone else
  • Positioning is a business decision; content follows from it

Make Content That Is Actually Useful

  • Generic content has been seen before and AI systems know the difference
  • Answer real, specific questions drawn from real customer language
  • Use support tickets, sales conversations, and reviews to find genuine query patterns
  • Depth and clarity matter more than volume

Build Pages AI Can Point To

  • Honest comparison pages with real trade-offs are high-value citation targets
  • How-to content with clear steps maps naturally to how AI systems construct answers
  • Use-case content tied to specific situations outperforms generic features pages
  • JSON-LD structured data tells AI systems what your content actually is; without it, you are asking them to guess

Use Real Visuals

  • Genuine screenshots, recordings, and product demos outperform stock imagery
  • Real visuals are a trust signal for both AI systems and the people who evaluate them
  • Generic imagery makes your brand indistinguishable from the competition

Expand Beyond Your Own Site

  • AI systems read third-party mentions, reviews, press coverage, and creator content
  • PR is part of your AI visibility strategy, not separate from it
  • Target the publications, communities, and creators your audience already trusts
  • Earn mentions through genuine relationships and content worth referencing

The Underlying Logic

  • Clarity, depth, and usefulness level the playing field against bigger competitors
  • EEAT principles have not changed; the consequences of ignoring them have
  • AI citation is not a budget competition; it is a quality and structure competition
  • Businesses that build this foundation now will have a meaningful head start

Search: AI Ate Your Traffic. Now What?

| admin

The new rules of search: zero click, multi-platform and brand led

Search used to work in a simple way. You typed a question, Google gave you a list of links, you clicked one, and you landed on a website. That website might belong to a business trying to sell you something, answer your question, or both. The click was the whole point.

That model is breaking down.

What has actually changed

AI tools like Google’s AI Overviews, ChatGPT, Perplexity and others now answer your question directly, on the spot, without sending you anywhere. You ask “what’s the best way to insulate a loft?” and you get a full answer, right there, no clicking required.

For the person asking the question, this is brilliant – hence the habit change. For the business whose website used to get that click, it is a serious problem.

The visit to your site was the start of everything commercially useful: someone reads your content, likes what they see, fills in your contact form, or buys something. If the AI answers the question before they ever reach you, that visit never happens. No visit, no lead. No lead, no sale.

This is what people mean by zero-click search. The question gets answered, but nobody goes anywhere.

It is not just Google any more

On top of this, search is no longer one place. While it never was exactly, Google has been preeminent for two decades. People are now asking questions on ChatGPT, Perplexity, Gemini, Apple Intelligence, Microsoft Copilot, and a growing pile of AI tools built into browsers, phones and apps. Each of these has its own way of finding information and deciding who to credit.

For years, SEO meant optimising your website for Google. One engine, one rulebook, broadly understood. That is still relevant, but it is no longer the whole picture. You now need your content to be findable, usable and citable by a range of AI systems that each work slightly differently.

What gets a brand or business cited

When an AI app does mention a source, it is not random. These systems consistently favour content with specific characteristics;

Concrete, specific facts. AI tools like content that contains clear, direct, checkable information. A claim like “our service covers the South East” is not citable. A claim like “we reduced average page load time by 40% across 12 client sites” is. The AI needs something it can pull out and attribute accurately.

Clear question and answer structure. Content written around real questions, with direct answers, tends to do well. This is partly because AI models are trained on that kind of content, so they recognise and trust the pattern. If your content dances around a question rather than answering it plainly, it tends to get skipped.

A brand the AI actually knows. This one surprises people. If the AI has no clear sense of who you are as a business, it will not name you. That means consistent naming across your website, your social profiles, your directory listings and any press coverage. The more places your brand appears in a coherent, consistent way, the more likely the AI is to recognise you as a real, trustworthy entity worth citing.

Other people saying the same thing. AI systems are more confident citing a claim if they have seen it backed up in multiple places independently. Coverage in trade press, customer reviews, expert mentions, and third-party references all help. If only your own website makes a particular claim, the AI may quietly ignore it.

A website the AI can actually read. This is straightforward technical hygiene. If your site is slow, badly structured, blocks crawlers, or is missing basic schema markup, AI systems may simply never see it. You cannot be cited if you cannot be found.

Why brand now matters more than ranking

Put all of this together and you get to the uncomfortable truth: being on page one of Google is no longer enough on its own.

What matters now is whether the AI knows your brand, trusts your content, and considers you worth mentioning when a relevant question comes up. That is a different challenge from traditional SEO, and it sits much closer to how you build a brand reputation than how you build a link profile.

The businesses that will do well in AI search are the ones that own a clear point of view in their field, produce specific and useful content regularly, and build enough presence across the web that the AI has plenty of good reasons to name them.

The click is no longer guaranteed. The mention is the new metric. Getting ready for that shift is the work.

Steve Coulter, State Of The Art Digital – May 2026

Search: Non-Commodity Content. What?

| admin

This is not a drill.

I make no apologies that this is a long read, but a vital one for all business owners.

The rise of AI search summaries and your highly probable non-inclusion is an existential travesty that your present agency has not flagged for you. A major problem compounded by Googles’ latest May 2026 ‘AI Optimisation Guidance’ update that totally prioritises Non-Commodity content. What? I hear you say.

Grab a coffee and learn what will make your business not only preeminent in search, but taking an unassailable early adopter advantage. I’ve used the example of Estate Agency (Real Estate Agency for US friends).

Thank me later.

The End of Commodity Content: Why Estate Agents (& All Other Businesses) Must Build Proprietary Knowledge Assets.

The strategic advantage now available to hyperlocal businesses across every sector is unprecedented. Whether you operate as an estate agent in Tunbridge Wells, a dental practice in Harrogate, or a veterinary surgery in Exeter, the competitive landscape in your immediate geography is about to be reset. Google’s May 2026 shift to answer-optimised search means that the first business in each town to build substantive, non-commodity content will dominate AI citations for their sector. The locksmith, solicitor, or accountant who documents genuine local expertise in structured, citable form will appear in AI Overviews while competitors remain invisible. This is not incremental advantage. This is first-mover monopoly in local search visibility. Every hyperlocal business category in every town is currently wide open: whoever moves first and builds the knowledge assets wins the territory. For estate agencies, dental practices, veterinary surgeons, and every other geographically bound service business, the question is whether you recognise this as the fundamental strategic opportunity it represents, or whether you let a competitor claim it while you continue publishing the same generic content as everyone else.

Google’s announcement in May 2026 represents the most significant shift in search behaviour since the introduction of mobile-first indexing. The mandate is unambiguous: AI Overviews and AI Mode will prioritise answer engines over traditional link farms, and websites that cannot demonstrate genuine expertise through original, substantive content will lose visibility entirely.

For estate agencies, this creates an immediate strategic problem. Most agency websites currently operate as variations on the same template: property listings fed from the same CML data, area guides plagiarised from Wikipedia, service pages that promise “expert local knowledge” without providing any, and blog content recycled from national property portals. None of this will survive contact with generative engine optimisation.

The technical term for what Google now penalises is commodity content: information that exists in functionally identical form across multiple domains. If your area guide for Cheltenham could be republished word-for-word as an area guide for Harrogate by changing only the place names, it has no value to a language model trying to synthesise authoritative answers. Google’s AI will cite the original source or the most comprehensive treatment, not the fifteenth derivative version.

What answer optimisation actually means

Answer engines work by parsing structured content to construct responses to natural language queries. When someone asks “what should I know before buying a Victorian terrace in Clifton”, the AI doesn’t return ten blue links. It synthesises an answer from multiple sources, citing only those that contribute novel, specific, verifiable information.

Traditional SEO optimised for ranking factors: keyword density, backlink profiles, domain authority. GEO optimises for citability: is your content substantive enough to be quoted as a source? Does it contain specific claims that can be verified? Does it offer information that cannot be derived from other published sources?

For estate agencies, this requires a fundamental shift from marketing copy to knowledge publishing. The question is no longer “how do I rank for ‘estate agents Bristol'” but “what do I know about property in Bristol that nobody else has documented?”

The proprietary knowledge problem

Most agencies possess substantial proprietary knowledge. The negotiator who has handled three generations of the same family understands inheritance patterns and family property decisions. The valuer who has appraised every house type in the town knows which streets command premiums and why. The lettings manager who has placed five hundred tenants understands seasonal demand patterns and rental price elasticity.

Almost none of this knowledge is published. It sits in email threads, verbal exchanges, and institutional memory. Meanwhile, the agency website publishes generic content about “our commitment to service excellence” and “comprehensive local knowledge” without ever demonstrating what that knowledge comprises.

The challenge for now and ongoing is making proprietary knowledge externally visible in structured, citable form. This means original research, original photography, original data analysis, and original testimony from sources who cannot be replicated.

Examples of non-commodity content that works

Commission your maintenance contractors to document common issues by property age and type. Get the plumber to explain what causes damp in 1930s semis versus Victorian terraces, which boiler brands fail most frequently, what actually needs replacing versus what can be repaired. This is knowledge derived from hundreds of callouts across your patch. Nobody else has it in this form.

Analyse your own transaction data to identify patterns invisible in national statistics. Document average void periods by property type, most common reasons for offer rejection, the actual gap between asking price and achieved price across different streets. Publish the findings with specific numbers, specific locations, specific time periods. This is proprietary data that cannot be sourced elsewhere.

Create measurement guides showing what physically fits in local property types. Which Victorian terraces can accommodate a standard three-seater sofa up the stairs, typical room dimensions in Edwardian semis for furniture planning, whether king-size beds fit in second bedrooms of common house types. This requires access to hundreds of properties and tedious documentation work. It is also exactly the kind of specific, practical information that answer engines will cite.

Interview long-standing residents about lived experience in the area. The family who have been in the same street for forty years can explain how the high street has changed, which local amenities have closed or opened, what the community rhythm actually feels like. These testimonials should be specific: names, dates, verifiable details. Not “I love living here” but “we moved here in 1987 when the factory was still operating, the high street had three butchers then, now it’s all coffee shops but the bakery on Crown Street is still the same family”.

Document the informal knowledge that demonstrates embeddedness. Which builder works regularly in the conservation area and understands the planning constraints, where you can actually get a plumber at short notice, the tree surgeon who knows the local authority’s approval process. This is concierge-level information that proves you are part of the community fabric rather than simply claiming it.

The controversy trap

The instinct when pursuing distinctive content is to reach for controversy: planning disputes, flooding risks, infrastructure problems, local political divisions. This demonstrates knowledge but introduces doubt at precisely the moment you need to build confidence.

An estate agency exists to facilitate transactions. Content that raises problems without resolving them creates friction in the buying decision. The goal is to prove local expertise while reducing perceived risk, not increasing it.

Better to focus on practical knowledge that helps buyers and sellers make informed decisions: seasonal patterns in your specific market, which streets have the strongest demand from families versus young professionals, what improvements actually increase sale prices based on your transaction data, which solicitors and mortgage brokers your clients report back as being efficient.

The content should answer questions people are genuinely asking but struggling to get answers to. Not “why choose us” but “what do we know that helps you”. The former is marketing. The latter is knowledge publishing.

Why this matters now

Google’s May 2026 mandate is not optional. Nor is your future inclusion in the results of AI Apps like Chat GPT, Claude, Gemini and Perplexity. Agencies and any business that continue to rely on commodity content will lose visibility as AI Overviews and AI Mode become the dominant search interface. The traffic that currently arrives via traditional organic search will increasingly be answered directly by the AI without a click-through. The so-called Zero Click phenomenon, or ‘Position Zero’.

The only websites that will retain visibility are those cited as sources in AI-generated answers. Citation requires original, substantive, structured content that contributes information unavailable elsewhere.

For agencies, this means treating content creation as knowledge asset development rather than marketing overhead. The investment required is significant: staff time to document expertise, original photography, data analysis, commissioned testimony. But the alternative is gradual invisibility as search behaviour shifts away from link-based results.

The agencies that will dominate local markets post-May 2026 are those that have built citable knowledge assets demonstrating genuine expertise. Not those with the best marketing copy, but those with the most substantive published evidence of what they actually know.

This is not a speculative trend. It is a structural shift already underway. The question is whether your agency treats it as an optional nice-to-have or as the fundamental precondition for future visibility.

For more information on how your business can capitalise on this paradigm shift please contact me.

From Your Correspondent: Google Might Be About To Widen The Pool

| admin

For over a decade, SEO has operated within a fixed constraint: Google’s deep learning ranking systems only evaluate the top 20–30 candidate pages because running neural networks on more results is too expensive. That number wasn’t chosen for quality reasons. It was set by hardware budgets and memory costs. Court testimony from Google’s VP of Search confirmed it, and now Google Research has published the algorithm that could remove the constraint. TurboQuant compresses vector representations by 4x whilst maintaining retrieval quality, making it economically viable to evaluate far larger candidate sets. When the ranking window widens, the rules change. Sites with strong content and structured data get a fair hearing against established players with dominant backlink profiles. The moat around incumbent rankings is about to shrink.

—

Google has historically ranked pages using a two-stage process that evaluates tens of thousands of candidates before applying deep learning (RankBrain, BERT) to just 20–30 finalists. This narrow window exists because running neural ranking on more pages is too expensive in compute and memory. That constraint was confirmed under oath by Google’s VP of Search, Pandu Nayak, during the DOJ antitrust trial.

Now the hardware economics are shifting. Google has published TurboQuant, a vector compression technique that reduces memory requirements by 4x whilst keeping retrieval quality high. CEO Sundar Pichai has acknowledged severe supply constraints on memory and foundry capacity, but TurboQuant addresses exactly that bottleneck by making retrieval indexing “virtually free” and reducing memory load per vector dramatically.

If deployed, TurboQuant lets Google evaluate a much larger candidate set before final ranking without adding hardware cost. The 20–30 page window was never a design decision. It was a budget ceiling. When the ceiling lifts, the entire competitive surface changes.

Why widening the search pool is good news

A wider candidate set levels the playing field. Under the current constraint, strong content on smaller or newer sites often never reaches the deep learning ranking layer because it gets culled in early retrieval stages dominated by classical signals like domain authority and link equity. The top 20–30 slots tend to go to established players with robust backlink profiles, not necessarily the pages with the best answers.

When Google can afford to evaluate 100 or 200 candidates instead of 20, retrieval-ready content gets a fair hearing. Pages with clear, citable claims, strong entity associations and semantic coherence can enter the ranking window even without legacy domain authority. Sites that have invested in content quality and structured information rather than link-building arms races get a shot they didn’t have before. The moat around incumbent positions shrinks.

For SMEs, local businesses and specialist publishers without big backlink budgets, this matters. If your page is genuinely retrieval-friendly (meaning AI systems can extract, verify and cite it), you’re now competing on content merit in a larger pool rather than being filtered out before ranking even starts. The game shifts from “can I outrank these 20 entrenched sites” to “can I be one of the 100 or 200 pages Google considers worth evaluating”. That’s a much more achievable threshold for quality content.

In practical terms for your consultancy clients: automotive retailers and estate agents with well-structured, citation-ready content (clean JSON-LD, strong NAP consistency, clear expertise signals) will have a better chance of appearing in AI-mediated results and wider ranking windows than they do now, where they’re often squeezed out by aggregator sites with stronger link profiles.

The shift favours signal over legacy authority.

COMMENTARY: It’s The Ecosystem Stupid.

| admin

Google Search AI Optimisation: Practical Guide for SEO & GEO Experts

Core principle

Google’s AI-powered search features (AI Overviews, formerly SGE) work fundamentally the same way as traditional Search: they rank and surface content from the web index using established quality signals. The same content that ranks well organically can appear in AI summaries.

What you need to do

1. Stick to established SEO fundamentals

Quality content that follows Google’s existing guidance will perform in AI features. No separate optimisation track is required. Focus on:

  • E-E-A-T signals: Demonstrate expertise, experience, authority and trustworthiness
  • Helpful content: Write for people, not algorithms
  • Technical foundations: Fast loading, mobile-friendly, crawlable architecture
  • Structured data: Use schema markup where relevant (though not a ranking factor, it helps Google understand context)

2. Understand how AI Overviews select content

AI Overviews pull from multiple high-quality sources to provide comprehensive answers. Your content is more likely to appear when:

  • It directly answers specific queries with clear, authoritative information
  • It ranks well in traditional search results for related queries
  • It demonstrates topical authority and expertise
  • It provides unique insights or perspectives not widely available elsewhere

3. Monitor performance differently

Track visibility using:

  • Google Search Console: Check impressions and clicks from AI Overview features (filter by search appearance type)
  • Traditional ranking data: Strong organic rankings remain the foundation
  • Click-through patterns: AI Overviews may reduce clicks for simple informational queries but can drive qualified traffic for complex topics

4. Optimise for citation-worthy content

Make your content more likely to be referenced:

  • Clear, factual statements: AI systems favour unambiguous information
  • Proper sourcing: Cite your own sources to establish credibility
  • Logical structure: Use headings, lists and clear paragraph breaks
  • Comprehensive coverage: Answer the full question, including related follow-ups
  • Unique data or insights: Original research, case studies or expert analysis stand out

5. Don’t try to game the system

Avoid tactics that attempt to manipulate AI features:

  • Writing specifically “for AI” rather than users
  • Keyword stuffing or unnatural phrasing
  • Creating thin content designed only to appear in summaries
  • Hiding text or using deceptive structured data

What doesn’t change

  • Quality over quantity: One excellent resource beats ten mediocre ones
  • User intent matters: Match content to what searchers actually need
  • Links still count: Authoritative backlinks remain a trust signal
  • Regular updates: Fresh, current information performs better for time-sensitive topics

What to watch

  • AI Overviews appear more frequently for:
    • Complex queries requiring synthesis from multiple sources
    • Questions where context and nuance matter
    • Topics where users benefit from seeing multiple perspectives
  • They appear less often for:
    • Simple navigational queries
    • Searches with clear commercial intent
    • YMYL (Your Money Your Life) topics requiring extreme caution

Measuring success

Success in AI features correlates with traditional SEO metrics:

  • Strong organic rankings (especially position 1-10)
  • High engagement metrics (time on page, scroll depth)
  • Topical authority (ranking for multiple related queries)
  • Quality backlink profile
  • Positive user behaviour signals

The single most important point: if your content ranks well and serves users effectively, it will appear in AI features when appropriate. There is no separate optimisation playbook.

Reality Check: Citation requirements for other LLM platforms

The above applies specifically to Google Search AI features. For citation and attribution in standalone LLM applications (ChatGPT, Perplexity, Gemini, Claude and similar), different factors apply:

Critical differences from Google

  1. No crawling schedule: LLMs access content through various methods (web search tools, direct fetches, training data cutoffs) with no predictable crawl pattern
  2. No ranking algorithm: There is no equivalent to PageRank or traditional ranking factors. Citation depends on relevance matching and content quality within the specific query context
  3. Inconsistent source attribution: Some platforms cite sources reliably (Perplexity, ChatGPT with search), others may reference content without formal attribution (Claude’s training data, Gemini’s knowledge base)

What increases citation likelihood across LLM platforms

Content characteristics that perform well:

  • Authoritative, factual content: Primary sources, original research, verified data
  • Clear, structured writing: LLMs parse well-organised content more effectively
  • Comprehensive topic coverage: In-depth resources that fully explore a subject
  • Recency: For search-enabled LLMs, recently published or updated content has an advantage
  • Quotable insights: Distinctive expert perspectives or unique data points that stand out
  • Accessible formatting: Clean HTML, proper semantic structure, readable without JavaScript

Technical factors:

  • Open access: Content behind paywalls or login walls is less likely to be cited
  • Crawlability: Standard robots.txt permissions (though some LLM providers may ignore these)
  • Fast loading: Some LLM search tools time out on slow sites
  • Mobile-friendly: Many LLM tools fetch mobile versions
  • Clear metadata: Title tags, meta descriptions and schema help LLMs understand context

Platform-specific considerations

ChatGPT (with web browsing)

  • Cites sources when using Bing search integration
  • Favours high-authority domains and recent content
  • Often pulls from news sites, academic sources and established publications
  • May quote directly with attribution when relevant

Perplexity

  • Most citation-focused of the LLM platforms
  • Provides numbered source references for most factual claims
  • Balances recency with authority
  • Particularly good at surfacing niche expert content if it ranks well

Gemini

  • Integrated with Google Search infrastructure
  • Similar source preferences to Google AI Overviews
  • Less consistent with citation formatting
  • Favours Google-indexed content

Claude

  • Training data cutoff means no access to recent content without web search
  • When web search is enabled, cites sources for factual claims
  • Prioritises authoritative, well-structured content
  • Less likely to cite unless directly relevant to query

What you cannot control

Unlike Google Search, you cannot:

  • Track when or how often you are cited by LLMs
  • Optimise specifically for particular platforms
  • Block individual LLM crawlers while allowing others
  • A/B test content for LLM citation rates
  • Measure referral traffic from LLM citations (except Perplexity, which passes some referrer data)

Practical approach for multi-platform visibility

  1. Optimise for Google first: Strong performance in Google Search increases likelihood of LLM citation
  2. Publish openly: Paywalls and registration requirements reduce LLM visibility
  3. Focus on expertise: LLMs preferentially cite recognised authorities and primary sources
  4. Structure clearly: Use semantic HTML, clear headings and logical content hierarchy
  5. Update regularly: For search-enabled LLMs, fresh content has an edge
  6. Create citation-worthy content: Unique data, original research and expert analysis stand out across platforms

The fundamental truth

No amount of technical optimisation will make poor content citation-worthy in LLM responses. The same principle applies here as with Google: create genuinely valuable, authoritative content that serves users, and citations will follow naturally.

The best strategy remains unchanged: be the best answer to the question being asked.

And breatheeeee…

If you’d like to focus on your business and leave this to a professional please contact me and let’s start a conversation.

Google’s Hubris: Why the Search Giant’s New AI Guide Proves It’s Already Lost

| admin

Google’s new May 2026 AI optimisation guide insists traditional SEO still works for AI Overviews, dismissing “GEO” as unnecessary.

But this misses the fundamental shift: ChatGPT, Claude, Perplexity and Gemini aren’t better search engines, they’re making search engines obsolete. Each AI platform rewards different content disciplines – structured authority, synthesis-friendly formats, factual density – precisely the foundations Google now downplays.

This is the Yahoo moment: a dominant platform trying to preserve its infrastructure while users are already trained to expect direct, conversational answers instead of ten blue links.

Optimise for Google if it drives traffic today, but the answer-optimised generation has moved on.

My new article tackles Google’s ego head on.

—

Google’s AI Optimisation Guide: A Masterclass in Missing the Point

Google published guidance last week on optimising for its AI Overviews and AI Mode. The subtext screams louder than the text: we’re worried, but we’re not changing.

The document insists that established SEO practices remain foundational. Keep your content crawlable. Use semantic HTML. Avoid duplicate pages. Create unique viewpoints. All sensible. All true for Google’s infrastructure. All increasingly irrelevant to where the search behaviour is actually moving.

Because here’s what Google won’t tell you: ChatGPT, Claude, Perplexity, and Gemini itself are training a generation to bypass the search page entirely. These aren’t alternative interfaces to the same underlying system. They’re fundamentally different paradigms, each with distinct ranking signals, citation logic, and content preferences.

Google’s guide dismisses “GEO” as unnecessary terminology. It claims you don’t need special markup, content chunking, or AI-specific files. It frames everything as continuous with traditional SEO. That’s technically accurate for Google’s implementation because Google bolted generative AI onto 25-year-old link-ranking infrastructure. It’s RAG as retrofit, not redesign.

But step outside Google’s walled garden and the picture changes completely. Each AI platform rewards different disciplines:

Perplexity favours recency and structured factual density. Content that can be cleanly extracted and attributed performs. Verbose preambles don’t.

ChatGPT prioritises synthesis-friendly formats and clear conceptual frameworks. It will reconstruct your argument if you’ve made it well, but it won’t wade through keyword-stuffed commodity content to find it.

Claude (and I’m obviously biased working with it daily) responds to authoritative voice, logical structure, and evidence-based reasoning. It cites sources that demonstrate expertise, not just keyword coverage.

Gemini sits awkwardly between Google’s traditional ranking systems and genuine generative behaviour, trying to serve two masters.

The disciplines I’ve been writing about for months (structured information architecture, clear semantic relationships, evidence-based authority signals, format optimisation for extraction and synthesis) matter more in these environments, not less. Google’s guide explicitly downplays several of them, not because they’re ineffective, but because Google’s implementation doesn’t rely on them as heavily.

This is the Yahoo moment Google won’t acknowledge. Yahoo didn’t fail because it stopped being a good web directory. It failed because web directories became the wrong answer to how people wanted to find information. Google is a superb search engine. The question is whether “search engine” remains the right category.

The answer-optimised generation don’t want ten blue links. They don’t want to triangulate truth across multiple sources. They don’t want to perform the ritual of “searching.” They want the answer, with enough provenance to trust it, delivered conversationally.

Google’s guidance tells publishers to optimise for Google’s systems while those systems themselves face displacement. That’s not strategy. That’s hoping the paradigm holds.

Nothing lasts forever. Not Yahoo’s directory. Not Google’s page rank. The platforms training users to expect direct, synthesised, conversational answers aren’t building better search engines. They’re making search engines obsolete.

The smart move isn’t ignoring Google’s advice. It’s recognising what the advice reveals: a dominant platform trying to preserve its infrastructure while the ground shifts beneath it. Optimise for Google if Google drives your traffic today. But if you’re planning for tomorrow, understand that each AI platform has distinct requirements, and the foundations that matter increasingly aren’t the ones Google built its empire on.

The reformation isn’t coming. It’s here and the incumbent just published a guide explaining why everything’s fine, actually.