Home / AI search visibility
Visibility in the answer engines
AI search visibility, or GEO, is the work of making a company citable by the engines that answer instead of listing. As a GEO and AEO consultant, I install that visibility, I measure it, and I show you the tracking month after month.
AI search visibility covers the techniques that make content extractable and citable by the answer engines: ChatGPT, Perplexity, Gemini and the AI Overviews of Google. It rests on three pillars: a direct answer placed at the top of each section, clean structured data markup, and mentions of the brand on the third party sources that the models consult.
Unlike classic search optimisation, the goal is no longer to win a click from a list of links, but to be the source that the engine reuses in its answer.
01 / What changes
The three coexist. None replaces another, but the work required is not the same and the results are not measured with the same tools.
| Discipline | Objective | The main lever | How it is measured |
|---|---|---|---|
| SEO Search Engine Optimization |
Appear in the list of results and win the click. | Domain authority, inbound links, relevance of the page. | Average position, impressions and clicks in Search Console. |
| AEO Answer Engine Optimization |
Occupy the answer shown at the top of the page, before the results. | Extraction structure: direct answer, lists, tables, marked up questions. | Presence in featured snippets and in the related question blocks. |
| GEO Generative Engine Optimization |
Be the source that the model cites when it writes its answer. | Factual citability, consistency of the entities, presence on third party sources. | Manual tracking on a list of typical questions, engine by engine. |
A good ranking in Google does not guarantee being cited by a model.
The two systems do not select on the same criteria. A page can be first on its query and stay absent from every generative answer, because it is written to be read and not to be extracted.
02 / Method
A typical engagement runs in four stages. The first measurement is dated and kept: without a starting point, no progress can be demonstrated.
I build a list of fifteen to thirty questions that your customers really ask, in their own language. I put them to ChatGPT, Perplexity, Gemini and Google, then note for each one whether you are cited, mentioned without a link, or absent, and who holds the place instead of you.
I examine what stops your pages from being picked up: access of the AI crawlers in the robots.txt file, structured data that is duplicated or unsuitable, structure of the answers, coverage of the entities of your sector, freshness and verifiability of the claims.
I correct, then I have corrections made or I write, depending on the agreed scope: direct answers moved to the top of each section, questions and answers that are visible and marked up, markup cleaned and consistent from one page to the next, llms.txt file, plan of external mentions.
The same tracking, every month, compared with the starting point. Change in the number of citations per engine, pages that gain, pages that stall, and the priorities of the following month set from the figures.
Citations take weeks to appear.
The models do not recompute their sources continuously. A first movement usually shows after six to eight weeks, and it takes a full quarter to judge a trend. Anyone who promises you a result in a fortnight is not measuring.
03 / Deliverables
Four pieces, one of which serves you well beyond the engagement: the dated tracking sheet.
Document
GEO score and AEO score, page by page.
Table
Your typical questions, engine by engine, from the starting point to today.
Production
Markup, answer blocks, questions and answers, llms.txt.
And then
The pages to create in order to cover the questions where you are absent, ranked by the ratio between search intent and production effort.
This is the document that turns an audit into a work programme, whether you run it in house or hand it to me.
04 / References
Portfolio
Tourism and events group, in French and in English.
Measured result
On queries for corporate offsites, incentives and team building in Portugal.
Demonstration
I publish my own citation tracking, month by month.
05 / Questions
A GEO consultant makes a company citable by the answer engines, then measures that citability every month on a stable list of questions.
The work splits into three parts: the structure, that is making the pages extractable by a model; the technical side, that is the structured data markup and the access of the AI crawlers; and the authority, that is the presence of the brand on the third party sources that the models consult. GEO consulting differs from classic SEO consulting on that last point: it is no longer about winning links for a ranking algorithm, but about existing in the sources of a model that writes.
SEO targets the click from a list of links, AI search visibility targets the citation inside an answer written by a model.
The two share common foundations, a technically sound site and relevant content, but AI search visibility adds structuring work for the extraction and presence work on the third party sources that the models consult. You can be first on Google and absent from every generative answer.
Allow six to eight weeks for a first movement, and a full quarter to judge a trend.
The models do not re-evaluate their sources continuously, and some of them rely on a search index updated at its own pace. That is why I establish a dated baseline before any work: without a starting point, it is impossible to tell progress from variation.
Yes, by putting a fixed list of questions to the engines every month and noting who is cited.
There is no equivalent of Search Console for the generative engines. The reliable measurement remains the tracking of a stable list of questions, carried out under the same conditions from one month to the next. Some tracking tools are starting to offer this measurement automatically, as a complement to the manual tracking and not as a replacement.
No, in the vast majority of cases the fixes concern the existing content and the markup.
Most of the gain comes from precise moves: lifting the answer to the top of the section, adding questions and answers that are visible and marked up, cleaning up the structured data, checking that the AI crawlers are not blocked. A rebuild is only required if the architecture of the site prevents those fixes.
It is a recent convention, still unevenly followed, whose cost of implementation is nil.
This file placed at the root of the site describes in plain text what the company does and lists the reference pages. No engine guarantees that it is taken into account to this day. I deploy it because the effort is counted in minutes, but I never present it as a major lever, contrary to what is often written.
It always starts with a citability audit, which stands on its own if you want to stop there.
The audit includes the baseline measurement, the technical review and the prioritised plan. What follows is decided on reading it: either you run it in house with the plan as a roadmap, or I take the workstreams on, on a fixed fee or as monthly support.
06 / Recommendation
We already had a few websites and the will to rethink the strategy. Armel being the guy I trust, I called him. I could not believe he managed to outperform our previous ranking with a brand new strategy, implemented at a much faster pace than I would ever expect. Ranked number one in a month for several queries.
Get in touch
Ask ChatGPT the question that your best customer would ask before finding you. If your name does not come up, write to me: I will tell you why, and what should change first.