I work with companies anywhere in the world, and specialise in industrial and manufacturing businesses selling into supply chains, from local to global. Buyers, engineers and project managers put shortlist questions to ChatGPT, Perplexity, Gemini and Google's AI Overviews before they contact anyone, and the companies named are the ones those systems can identify, retrieve and attribute. I make businesses visible to AI search, then help you enter new markets and close what comes in.
The check runs on your company, not on mine. You get a recorded walkthrough back.
Answer engines do not hand back ten blue links. They return a short list — usually a handful of named companies with citations attached — and most buyers act on it without scrolling further.
If your name is not in that list, you are not competing on price or capability. You are not in the conversation at all, and no amount of website traffic tells you it is happening. That is an observation about how the format works, not a prediction about your business — which is exactly why the first thing I do is go and look at your actual answers.
who are the best suppliers for [what you sell] in [your market]?
Based on the sources I can find, these are the ones that come up most often:
Illustration of the answer format — not a captured result. The free check produces the real version, with your name, your competitors and the date it was run.
Many modern sites build their pages inside the visitor's browser rather than on the server. A human browser runs that code and sees a finished page. AI crawlers do not run it, so what they receive is whatever existed before the code executed — which on some sites is a navigation bar and nothing else.
This is the most concrete single reason a company is absent from AI answers, and it is invisible from the inside: your analytics look normal, because the people you can measure are the ones whose browsers work. Whether it applies to you takes about a minute to establish, and it is the first thing the free check reports.
Vercel and MERJ tracked over 500 million GPTBot requests and reported in December 2024 that they found no evidence of JavaScript ever being executed: GPTBot requested JavaScript files 11.5% of the time and ClaudeBot 23.8%, and neither ran them. Google's Gemini and Applebot are the exceptions, because both reuse rendering infrastructure built for classic search.
Answer engine optimisation (AEO) is the practice of making a company's information retrievable and citable by AI systems that answer questions directly — ChatGPT, Perplexity, Google's AI Overviews, Gemini and Claude — rather than returning a page of links. It matters now because those systems assemble an answer from a handful of retrieved passages and attach citations to them, so a business that is not retrievable in passage form is absent from the answer even when its website ranks well in classic search. Generative engine optimisation (GEO) describes the same discipline; the two terms are used interchangeably by buyers.
It is not a replacement for search engine optimisation. Answer engines still retrieve heavily from indexed web content, so a site that classic crawlers cannot read is a site answer engines cannot cite either. The two disciplines sit on top of each other, in that order.
Three published findings, each with the study, the date and the sample size attached, so you can check every one of them yourself. Together they describe a shift that is already measurable rather than a prediction about where things might go.
of agencies report rising client demand for AEO and SEO for AI-driven search — the number one new service requested in 2026, ahead of paid ads (55%) and short-form video (51%).
AgencyAnalytics, 2026 Marketing Agency Benchmarks Report — 494 agency professionals surveyed, February–April 2026.
of AI citations came from sources a brand can manage or strongly influence — 44% first-party websites, 42% listings and profiles.
Yext, October 2025 — 6.8 million citations across 1.6 million AI answers from Gemini, OpenAI and Perplexity.
of pages cited in Google's AI Overviews also ranked in the top ten for the same query — down from 76% eight months earlier. Ranking and being cited are drifting apart.
Ahrefs, March 2026 — 863,000 keywords and 4 million AI Overview URLs, compared against the same study from July 2025.
Read those three together and the conclusion is obvious: demand is real, most of what gets cited is inside your control, and ranking alone no longer buys you a citation. The fourth constraint — whether an AI crawler can read your pages at all — is described above, because it is a mechanism rather than a number.
The first is what the name stands for — SEO, AEO and GEO are one discipline, not three, and it decides whether a buyer finds you at all. The other two decide what happens next: whether the market was worth entering, and whether the conversation becomes a signed order.
AEO, GEO and SEO
Make the business resolvable as an entity, make its pages retrievable as self-contained passages, and make the third-party sources engines cross-check say the right thing. Measured as answer share and citation rate across ChatGPT, Perplexity, Gemini, Claude and Google's AI Overviews.
Go-to-market strategy
Test whether a country or segment will actually buy before committing to a local hire. Segment definition, competitive reality, route to market and a live proposition test — ending in a hire, partner, persist or stop recommendation with the evidence behind it.
Win/loss, deal strategy
Discovery goes well, then the deal stalls inside the buying committee and nobody ever says no. Diagnose which stall it is, map who signs and who blocks, and reduce the buyer's downside rather than the price.
Being cited is not luck and it is not a content volume problem. It is a sequence: an engine has to identify you, find you, be able to quote you, and find the same story corroborated elsewhere. Skip a stage and the ones above it stop compounding. The whole method is published, in enough detail to interrogate on a call — how a prompt set is built, what counts as a citation, how it gets measured.
Before an engine can recommend a company it has to know which company you are. That means a single canonical description, a complete and truthful set of profile links, consistent naming everywhere the business appears, and resolving collisions with similarly named entities. Skip this and every later stage strengthens someone else's brand.
Keywords are not prompts. I build a fixed set of the questions your buyers put to an assistant — comparisons, shortlists, specification and procurement questions — written down, versioned and re-run unchanged, so results are comparable month to month rather than cherry-picked.
Retrieval works on passages, not pages. Sections lead with a direct answer, stay self-contained, define their own terms, and sit under headings phrased the way the question is asked. Structured data declares what the page is. Server-rendered HTML makes sure the content exists for crawlers that do not run JavaScript.
First-party pages are not the whole picture. Yext's October 2025 analysis of 6.8 million citations found 42% came from listings and profiles against 44% from first-party sites. Directories, profiles and third-party pages get the same care as the website, because engines cross-check across them.
The prompt set is re-run on a schedule across the engines your buyers actually use. What gets reported is how often you were named, how often you were cited, which source got the citation, and who was named instead. Traffic is not the metric; presence in the answer is.
Conversions are the metric clients say they care about most — 44% of agencies name it, against 6% for traffic (AgencyAnalytics, 2026, n=494). AI visibility does not report itself in an analytics dashboard, so the chain from citation to enquiry has to be built deliberately. This is how I build it.
| Metric | How it is produced | What it tells you |
|---|---|---|
| Answer share | Share of the fixed prompt set where your company is named at all, per engine | Whether you are in the conversation |
| Citation rate | Share of those answers where a source of yours is the attached citation | Whether you are the source or just a mention |
| Cited surface | Which URL or profile earned the citation — site, listing, directory, third party | Where to put the next hour of work |
| Competitive set | Who is named instead of you, and from which source | What the engine currently believes the category is |
| Assisted enquiries | Referral traffic from AI surfaces, plus a “how did you hear about us” field on your own enquiry form | The closest honest link to pipeline — self-reported attribution is imperfect and is reported as such |
ChatGPT, Perplexity, Google AI Overviews, Gemini and Claude, checked separately rather than blended into one score. They retrieve and cite differently, so a single average hides the thing you need to see.
One property of the technology governs how every number above should be read: answer engines are non-deterministic. The same prompt can return different answers on different days and to different users. That is why the prompt set is fixed and re-run on a schedule — a trend across repeated runs is evidence, a single screenshot is not.
AI visibility is probabilistic. There is no ranking factor to buy, no position to hold and no lever any vendor controls end to end. That is a property of the technology, and it is worth being precise about before you compare proposals.
That distinction — between the part that is engineering and the part that is probability — is the whole discipline. Anyone who collapses the two is either confused or selling. You are entitled to ask which, and to ask it of me.
No, but it depends on SEO. Search engine optimisation gets a page indexed and ranked; answer engine optimisation makes a specific passage on that page retrievable, quotable and attributable by a system that answers in prose. The overlap is real — an unindexable site fails both — but the unit of work is different: SEO optimises pages, AEO optimises passages and the entity behind them.
In practice, none. Answer engine optimisation and generative engine optimisation describe the same discipline, and the industry uses them interchangeably. GEO is the term used in the academic literature, notably the 2024 KDD paper by Aggarwal and colleagues at Princeton and IIT Delhi; AEO is more common among buyers. I keep separate pages for both because people search for both.
Ask what they measure. A rebranded SEO package reports rankings and traffic with an AI section bolted on; an actual AEO engagement reports answer share and citation rate against a written prompt set, per engine, and can tell you which specific source earned each citation. The second question worth asking is what they do about your listings and profiles, since Yext found in October 2025 that 42% of AI citations came from those rather than from first-party sites.
A fixed set of buyer questions is written down and re-run unchanged across ChatGPT, Perplexity, Gemini, Claude and Google's AI Overviews on a schedule. Each run records whether you were named, whether one of your sources was the citation, which source it was, and who was named instead. Because answer engines are non-deterministic, the evidence is the trend across repeated runs, never a single screenshot.
It depends on crawl and refresh cycles rather than on effort. Entity and structural changes have to be re-crawled and re-retrieved before they can affect an answer, so nothing moves the same week, and different engines pick changes up at different rates. That is precisely why the prompt set runs on a schedule: the timeline becomes an observation about your category rather than a promise about it.
No, and neither can anyone else. Answer engines choose sources at the moment they assemble a response, using retrieval systems their operators change without notice. What can be controlled is whether your business is identifiable, whether your content exists in a retrievable form, and whether the sources engines lean on say the right thing about you.
Usually not, but it depends on how your pages render. Vercel and MERJ reported in December 2024 that no major AI crawler executes JavaScript — GPTBot requested JavaScript files 11.5% of the time and ClaudeBot 23.8%, and neither ran them. If your content only appears after JavaScript runs, most of these systems never see it, and that is a build problem rather than a content one. The free check tells you which situation you are in.
Any company, anywhere in the world, that sells something a buyer researches before making contact. I specialise in industrial and manufacturing businesses selling into supply chains from local to global, because those are the categories where a buyer asks an assistant for a shortlist rather than clicking an advert. Engagements run in English. It is one person, so a small number run at a time.
Send me your company name and website. I run a set of buyer questions across ChatGPT, Perplexity, Gemini, Claude and Google's AI Overviews, and send back a recorded walkthrough of what came out: where you appeared, where you did not, who was named instead, and the two or three structural reasons why. No charge, no obligation, and you keep the findings whether or not you decide to hire me.
I run these myself, so there is a queue. Expect a few working days rather than an instant report.
