Month: January 2026

AI SEARCH: The Science Behind GEO

| admin

Generative Engine Optimisation Isn’t a Buzzword. It’s a Research Discipline.

For the past year or so, *Generative Engine Optimisation* (GEO) has been talked about as if it were simply “SEO for ChatGPT”. That shorthand does it a disservice.

“GEO isn’t a marketing fad dreamed up in a boardroom. It is grounded in serious academic work on how AI-powered search systems actually operate – and that distinction matters if you want your content to be visible in the next generation of search.”

The scientific foundation was laid in 2023 by researchers from Princeton, Georgia Tech, the Allen Institute for AI, and IIT Delhi, in a paper later accepted to KDD 2024. This wasn’t speculative thought leadership; it was empirical research into how large language models retrieve, synthesise, and present information.

From Search Engines to Generative Engines

The research introduces a critical shift in thinking: we are no longer optimising for *search engines*, but for what the authors call *generative engines*.

Traditional search engines retrieve and rank documents. Generative engines do something fundamentally different. They retrieve information from multiple sources, synthesise it using large language models, and generate an original response – often with inline attribution.

That difference breaks many of the assumptions SEO has relied on for two decades.

GEO sits at the intersection of computational linguistics, cognitive science, and machine learning. Instead of focusing on keyword placement or ranking signals, it addresses how language models recognise patterns, how they use their context window, and how probability distributions shape what ultimately appears in a generated answer.

In short: it’s not about being number one on a list. It’s about being *included* in the synthesis.

How GEO Was Tested (And Why That Matters)

One of the strengths of the research is its methodology. The authors didn’t rely on anecdote or tool screenshots. They built GEO-bench: a large-scale benchmark of 10,000 queries spanning multiple domains, each tagged by intent, difficulty, domain, and expected answer format.

The experimental setup was deliberately realistic. First, top sources were fetched from Google Search. Then GPT-3.5-turbo was used to generate answers with inline citations, mirroring how generative search systems work in the wild.

Crucially, the researchers didn’t measure success using traditional rankings. They introduced new visibility metrics designed specifically for generative engines, including:

  • Position-adjusted word count (how prominently a source appears in generated responses)
  • Subjective impression scores, measuring relevance, influence, uniqueness, and likely user engagement

Using these metrics, they tested nine distinct optimisation methods.

The results were telling.

What Actually Improves Visibility in AI Search

Certain GEO techniques consistently outperformed others. Adding clear citations, quotations, and concrete statistics increased visibility in generated responses by as much as 40 per cent.

Meanwhile, many familiar SEO habits barely moved the needle. Keyword stuffing, in particular, showed minimal impact on generative engines – a finding that should give pause to anyone still writing for algorithms rather than understanding.

The research also highlights that GEO is not one-size-fits-all. Effectiveness varies by domain:

  • An authoritative, declarative tone works best for debate and historical topics
  • Citation-rich content performs strongest for factual queries
  • Structural clarity matters more than keyword density

This reinforces an uncomfortable truth for some marketers: optimising for AI means writing better, not trickier.

Why This Changes Content Strategy

GEO forces a rethink of what “optimised” content looks like. If your material can’t be easily understood, trusted, and recomposed by a language model, it risks invisibility – regardless of how well it ranks today.

“What the research makes clear is that generative visibility is earned through clarity, evidence, and structure. AI systems reward content that behaves like a good academic source or a solid piece of journalism: well-sourced, precise, and unambiguous.”

That may feel less glamorous than chasing hacks. But it’s a more durable advantage.

As generative engines continue to replace blue links with synthesised answers, GEO will stop being a niche concern and become a baseline competency. Those who treat it as a discipline – grounded in how these systems actually work – will be the ones whose voices are carried forward.

The rest will simply be paraphrased out of existence.

AI SEARCH: Red Alert. GEO Is Now Critical

| admin

For more than two decades, organic search followed a broadly predictable pattern. Rank higher, earn more clicks. Position one hoovered up attention, position two fought over the scraps, and by page two you were effectively invisible. Entire SEO strategies, pricing models and business forecasts were built on that curve.

In 2025, that curve has broken.

The widespread rollout of Google’s AI Overviews has fundamentally altered how users interact with search results. The most important statistic to understand this shift is not impressions, not rankings, and not even traffic. It is click-through rate by position.

And the change is not subtle.

The 2025 CTR Shock

Multiple large-scale studies now show that when an AI Overview appears, organic click-through rates drop sharply at the very top of the page.

Position one, historically responsible for roughly 27 to 30 percent of clicks, now often sees figures closer to 18 to 20 percent when an AI Overview is present. Position two has been hit even harder, with CTR reductions approaching 40 percent in some verticals. Positions three to five also decline, though less dramatically.

This is not seasonal noise or algorithmic wobble. It is structural.

“The reason is simple. Users are no longer starting their journey with organic listings. They are starting with a machine-written synthesis that sits above everything else.”

For many informational queries, the search ends there.

The Rise of the No-Click Result

AI Overviews represent the most aggressive expansion of the zero-click search model Google has ever deployed. Featured snippets were short. Knowledge panels were limited. AI Overviews are comprehensive, contextual and designed to resolve intent directly on the results page.

“In 2025, a majority of informational searches that trigger an AI Overview now result in no click at all.”

This matters because it breaks a long-standing assumption in digital strategy: that visibility inevitably leads to traffic. It no longer does.

A page can rank first, be technically sound, well written, and perfectly aligned with search intent, and still receive a fraction of the traffic it would have earned two years ago.

Why Lower Rankings Are Not the Answer

Some commentators have pointed out that positions six to ten sometimes see a relative increase in CTR when AI Overviews are present. This is true, but it is also misleading.

Those positions are benefiting from a smaller group of users who scroll deliberately to validate or explore sources after reading the summary. They are not outperforming top positions in absolute terms, and they are not a growth strategy.

This is not a reshuffling of clicks. It is a contraction of them.

The Real Divide in 2025 Search

The meaningful distinction in modern search is no longer between page one and page two. It is between content that is cited by AI systems and content that is merely indexed.

Being cited inside an AI Overview changes the equation. It restores relevance, trust and visibility at the point where the user’s attention actually is. It turns a passive summary into a gateway rather than a dead end.

Businesses that are cited consistently tend to see stronger branded searches, higher downstream engagement, and better conversion quality, even if raw organic traffic volumes are lower than historical peaks.

Businesses that are not cited experience something worse than a ranking drop. They experience quiet irrelevance.

Why Traditional SEO Is Now Failing Businesses

“In 2026 most SEO strategies are still built for a search landscape that no longer exists. They optimise for rankings rather than references, keywords rather than concepts, and pages rather than entities.”

AI systems do not think in keywords. They synthesise from sources they consider authoritative, current, structured and reliable. If your content is not written, structured and positioned to be used as a source, it is invisible to the most important layer of modern search.

This is why many businesses report stable rankings alongside falling traffic and weakening lead quality. The strategy is technically succeeding while commercially failing.

The Cost of Inaction

Choosing not to adapt is still a choice, but it is an expensive one.

If your content is not being cited, you are training AI systems to answer questions without you. Every un-cited article reinforces competitors as default sources. Every missed summary compounds future invisibility.

In 2025, search visibility compounds in two directions. Upwards if you are referenced. Downwards if you are not.

The Strategic Shift Required

The implication for business is clear.

“SEO is no longer about chasing clicks. It is about earning inclusion in machine-generated answers. That requires a shift toward Generative Engine Optimisation, whether or not that label is used internally.”

Content must demonstrate expertise clearly, answer questions directly, and be structured in ways AI systems can parse, trust and reuse. Authority signals matter more than ever. So does clarity, accuracy and topical depth.

Ranking still matters, but it is no longer the end goal. Being used as a source is.

The Bottom Line

The dramatic change in organic CTR by position is not a temporary anomaly. It is the clearest measurable signal that search behaviour has crossed a threshold.

Businesses that continue to optimise as if blue links are the primary interface are optimising for the past. Businesses that understand how AI systems select, summarise and cite sources are building visibility where it actually exists.

In 2025, search success is not about being first on the page. It is about being present in the answer.

LINKEDIN: The Logic Behind The 2026 Algorithm Pt.4

| admin

As LinkedIn shifts from a feed-driven model to a retrieval-based system, older content is no longer obsolete but conditional. Posts from previous years can re-enter circulation when present-day relevance reactivates them. This final piece examines how the platform now treats past work as dormant knowledge, and why coherence over time has become a decisive advantage.


Part 4: The Past Is Not Archived. It’s Dormant.

One of the stranger side effects of LinkedIn’s new identity is the sudden reappearance of the past.

Posts from 2024. Threads from 2025. Ideas that barely registered at the time drifting back into view, sometimes years later, as if the platform has developed a memory and decided it’s finally ready to use it.

Most people assume this is nostalgia, or randomness, or some minor quirk of the feed.

It isn’t.

What’s happening is reactivation, and it is one of the clearest signs that LinkedIn now behaves less like a social network and more like an answer engine.

Feeds forget.
Knowledge systems retrieve.

The old LinkedIn treated content as disposable. Once the moment passed, the post was effectively dead. The new system treats content as conditional. Dormant, not deleted. Waiting for a reason to matter again.

And that reason is always present-day relevance.

When an older post is commented on, shared with context, or even quietly rediscovered via profile exploration, it isn’t judged by the rules of the year it was written. It is evaluated by the rules of now. If it holds together and if it still answers a professional question cleanly then it re-enters circulation.

This is exactly how AI answer engines work. They do not privilege freshness by default. They privilege usefulness. Time is only a problem if it introduces error. Otherwise, survival becomes proof.

The same logic now applies at the profile level. If your recent work reinforces a topic you were already writing about years ago, the algorithm treats that continuity as evidence. You are not changing direction; you are confirming identity.

Old posts stop being “old”.
They become supporting material.

This is why comments matter more than people realise. A thoughtful comment is not just participation. It is a retrieval event. It pulls your thinking – past and present – back into view. It reminds the system what you are associated with, and how long you’ve been associated with it.

The system is not looking for novelty.
It is looking for confirmation.

This also explains why some content is never revived. Shallow takes age badly. Trend-dependent posts collapse without their context. Engagement bait dies the moment the crowd moves on. Time doesn’t rescue weak structure; it exposes it.

But well-formed thinking ages differently. It doesn’t spike, but it doesn’t decay either. It waits.

For people who wrote properly before the platform knew how to reward it, this moment feels oddly belated. Work that once seemed under-performant now reads like pre-training data. Not because it was clever, but because it was complete.

The important shift here is psychological. If you still think of LinkedIn as a feed, you’ll keep trying to keep up. If you recognise it as a retrieval system, you start thinking in layers instead of moments.

You don’t rewrite your past.
You reference it.
You echo it.
You let it resurface when the present gives it a reason to.

This is not content recycling. It’s identity reinforcement.

The uncomfortable implication is that nothing you post is truly finished anymore. Every piece either becomes part of a growing body of work – or it quietly disqualifies itself from being remembered.

Which brings us to the real divide opening up on the platform.

Some people are still posting to be seen.
Others are posting to be recognised – now, later, or by systems that haven’t fully arrived yet.

The new LinkedIn doesn’t reward urgency.
It rewards coherence.

Coherence, once established, has a long memory.

LINKEDIN: The Logic Behind The 2026 Algorithm Pt.3

| admin

As LinkedIn’s algorithm converges with AI-mediated discovery, visibility is no longer the primary currency. What matters now is whether an individual’s thinking is stable, attributable, and reliable enough to stand in for them over time. This third piece of four for the New Year explores the implications of that shift, and why professional recognition is quietly replacing reach as the platform’s defining reward.

Part 3: From Posting to Permanence

This is the part that makes people uncomfortable, because it suggests the end of something.

The creator era on LinkedIn is quietly winding down. Not with a backlash, but with indifference. Performance without substance no longer compounds. Visibility without usefulness no longer sticks.

What replaces it isn’t silence. It’s reference.

LinkedIn is preparing for a world where professional insight is increasingly mediated by machines. Internal copilots. AI-driven search. Summaries of “what people who know about this think”. In that world, the platform doesn’t need louder voices. It needs reliable ones.

Which means content must be defensible. Contextually complete. Stable over time. Clearly attributable to someone who appears to know what they’re talking about — and to have known it for a while.

This is why older posts that were written properly are suddenly resurfacing. Not because the algorithm is sentimental, but because time is now a positive signal. Surviving without contradiction is a form of validation.

The great misunderstanding is that this is about reach. It isn’t. Reach is incidental. The real competition now is for recognition – by humans first, machines second, as someone whose thinking can safely stand in for them.

That’s why the LinkedIn algorithm and AI summary standards now look so similar. They are solving the same problem from opposite ends. One curates what professionals see. The other curates what professionals ask.

Both are ruthless about the same thing and that is; useless content does not deserve to persist.

The feed, as we knew it, is effectively dead.

What’s replacing it is slower, quieter, and far more consequential: a professional answer engine assembling itself in public.

Those who understand this will stop chasing attention and start building intellectual permanence.

The rest will keep posting – and wonder why nothing seems to last.