01AI Search
AI search optimisation: being cited when the answer replaces the click
AI search work from $900 a monthMeasured per platform, never blended into one score. Baseline first, then movement.
AI Search is Searlixa's umbrella programme for getting a brand cited, named and recommended inside AI-generated answers. It covers measurement, topical coverage, entity clarity and technical access across ChatGPT, Google AI Overviews and AI Mode, Gemini and Perplexity.
Quick answer
AI Search is the umbrella programme for getting a brand cited inside AI-generated answers. Four workstreams: measurement, topical coverage, entity clarity and technical access.
Most of the optimisation is search work done properly. What is genuinely new is the measurement, and the fact that platforms diverge enough that a blended score hides the finding.
Key takeaways
- Ranking is no longer sufficient. Top-10 share of AI Overview citations fell from roughly 76% to 38% between July 2025 and March 2026.
- But it still matters most. Top-three pages appeared in AI citations 98.9% of the time.
- Breadth per page beats page count. Pages serving 8+ related queries hit the AI top ten 43.3% of the time against 15.5% for single-query pages.
- We report each platform separately. The two most similar shared only 24.1% of cited pages.
- Entity clarity is usually the fastest-moving workstream, because correcting a wrong fact is cheaper than building authority.
- No llms.txt, no chunking, no AI-specific schema. Google names all three as unnecessary.
Definition
AI search optimisation is the practice of making a brand and its content retrievable, quotable and correctly attributed inside AI-generated answers. Most of the underlying work is search work. What differs is the measurement, and the fact that each platform retrieves differently.
What changed, in numbers
Between July 2025 and March 2026 the share of Google AI Overview citations coming from top-10 organic results fell from roughly 76% to 38%, according to Ahrefs. A separate 2026 analysis of 1,645 AI citations found 55.2% of cited pages were not traditional top-10 results.
Ranking has not stopped mattering. In that same study, pages holding traditional top-three positions appeared in AI citations 98.9% of the time. What has changed is that ranking is no longer sufficient, and that a page can be cited without ranking at all.
Verified Published third-party research. Sources and limitations are listed on our measurement methodology.
The four workstreams
Measurement
A frozen buyer prompt set sampled across each platform on a fixed cadence. Citation presence, share of voice against a named competitor list, and which of your pages get used. Delivered by the two-week baseline and continued through ongoing sampling. If you would rather run it in house, the method is published in full.
Topical coverage
Pages that serve many related queries get cited far more than pages that serve one. In the 2026 citation study, pages reused across eight or more queries reached the AI top ten 43.3% of the time, against 15.5% for single-query pages. Depth on existing pages usually beats publishing new ones, which is why our SEO work starts with consolidation rather than a content calendar.
Entity clarity
Models have to resolve who you are before they can recommend you. Consistent naming, explicit relationships between your brand, products, people and locations, and agreement between what your site says and what third-party sources say. Covered by entity optimisation, which is usually the fastest-moving part of a programme.
Technical access
Crawlers you have chosen to allow must be able to reach and render your content. That includes a deliberate robots policy separating search discovery from model training, which are different decisions. Covered by technical SEO.
Why we report each platform separately
The two most similar AI platforms in the 2026 study shared only 24.1% of the pages they cited. Every other pair overlapped less. A blended "AI visibility score" averages away the one thing you can act on: that you are strong on one surface and absent from another.
| Surface | Behaves like | Emphasis shifts to |
|---|---|---|
| Google AI Overviews / AI Mode | Grounded in Google's index and ranking systems | Core SEO strength; see generative engine optimisation |
| ChatGPT | Highly exclusive source selection; own retrieval | Answer structure and third-party corroboration |
| Perplexity | Citation-forward, source-list oriented | Extractable claims and clear attribution |
| Gemini | Mixed grounding across Google surfaces | Entity consistency across your whole footprint |
| Assistants generally | Answer instead of linking | Being the answer; see answer engine optimisation |
What a programme looks like in the first ninety days
Roughly, and it varies by what the baseline finds.
- Weeks 1–2. Prompt set agreed and frozen, competitor list agreed, baseline sampled across the platforms in scope. Nothing is recommended before this exists.
- Weeks 3–4. Findings walkthrough and a prioritised gap list. In most engagements the first item is not content. It is a crawler policy, a rendering problem or a stale third-party profile stating something wrong about you.
- Weeks 5–10. Technical blockers cleared, canonical fact set assembled and propagated, and the two or three pages already close to being cited deepened rather than new ones published.
- Weeks 11–13. First re-sample against the same frozen set. Too early to call a trend, early enough to see whether anything is moving in the right direction.
Where AI Search ends and SEO begins
Honestly: mostly they do not separate. Google published documentation in May 2026 stating that optimising for generative AI features is optimising for search, and named llms.txt files, content chunking and AI-specific schema as unnecessary.
We agree with that. The reason this programme exists as a distinct service is measurement, not tactics. SEO remains the base layer and the strongest single gateway into AI answers. Anyone selling you a wholly separate AI playbook is selling you most of an SEO retainer twice.
Common starting points
Most brands arrive with one of three problems. Absent from assistant answers, missing from AI Overviews despite ranking, or organic traffic falling with rankings intact. Each guide is written so you can diagnose it yourself before deciding whether to involve anyone.
02Scope
What this programme will not include
llms.txt optimisation, content fragmentation for machines, AI-specific schema packages, guaranteed placements in non-deterministic systems, or paid-for mentions. Each is either publicly dismissed by the platform it targets or named in Google's spam policies, which now apply explicitly to generative responses in Search.
The glossary says plainly which of these terms describe genuinely different things and which are synonyms being sold as separate services.
03Common questions
What buyers ask before committing
Is AI search optimisation different from SEO?
Mostly not. Google states that optimising for its generative features is optimising for search, and the underlying work overlaps heavily. What is genuinely different is measurement: nothing in Search Console tells you whether an assistant cites you, on which platform, for which prompt.
Which platform should we start with?
Whichever your buyers use, which is usually Google AI Overviews by exposure. If you sell into a technical audience, ChatGPT often matters more. The $99 audit answers this in three working days rather than you guessing and committing a quarter to the wrong surface.
How long before we see movement?
Correcting a wrong fact in a source a model already uses can show within weeks. Building authority where none exists takes months. On a new domain in a competitive category, twelve to eighteen months is the realistic planning assumption.
Do we need to stop doing SEO?
No, and it would be a mistake. Traditional top-three positions appeared in AI citations 98.9% of the time in 2026 research. Ranking is no longer sufficient, but it remains the strongest single gateway into AI answers.
Can you guarantee we will appear in AI answers?
No, and nobody honestly can. Generative answers are non-deterministic: the same prompt returns different sources on consecutive runs. What we commit to is method: a frozen prompt set, a fixed cadence, documented coding rules and a named owner.
What does an AI Search engagement actually produce?
A baseline you can check, a prioritised list of gaps ranked by commercial value, the work itself delivered with or by your team, and re-measurement against the same frozen prompt set. Plus the raw sampling data, which is yours regardless.
→What you are buying
AI search and AI visibility services: what is included
What an AI search agency should hand you is measurement first and content second, because you cannot prioritise a gap you have not sized.
- A prompt set built from your own sales language, agreed and then frozen
- Sampling across ChatGPT, Google AI Overviews, Gemini, Perplexity and Copilot
- Three runs per prompt, so variance is visible rather than hidden
- Named competitor share of voice against a list you agreed
- Description-accuracy coding on every response, not just presence
- Entity fixes implemented across your own properties and the sources that describe you
Who this is for
Brands that rank and are still absent from the answers buyers read, and brands being described wrongly by assistants.
Who it is not for
Anyone wanting a single AI visibility score. Two platforms shared 24.1% of cited pages in 2026 research; averaging them destroys the only finding worth having.
From $900 a month. Fifteen prompts across three platforms at the entry tier, 120 across five at Enterprise. SEO and AI visibility pricing is published in full.
The audit produces your baseline: the prompt set, the first sampling, and where you stand against the competitors you named. The $99 SEO and AI visibility audit comes back in three working days, and the fee comes off your first invoice if you continue.
04Next
Go deeper
GEO · AEO · Entity optimisation · Start with a baseline · See a worked example · Pricing
Sources
Every figure on this page comes from one of the following. Where a claim is ours rather than a source’s, it carries an evidence label instead.
- Rankability — State of AI Search, 1,645 AI citation observations across 31 topics, 2026. Source for the 24.1% cross-platform overlap, the 73.5% ChatGPT exclusive citation rate, the 22.1% vs 8.7% first-100-words finding and the 43.3% vs 15.5% query-breadth finding. Independent study
- Ahrefs — AI Overview citation analysis, March 2026. Source for the fall in top-10 organic share of citations from roughly 76% to 38%. Independent study
- Google Search Central — Optimizing your website for generative AI features (published 15 May 2026, updated 10 July 2026). Source for the llms.txt, chunking and AI-schema positions, and for the statement that generative features run on core ranking systems. Primary
- Searlixa sampling across client engagements. Any figure drawn from this carries an Observed label with its sample size and date. Our own
Ask an assistant
Hand this page to an assistant and ask it to summarise or challenge the argument. If it gets something wrong about us, that is itself a useful finding.