What is AI visibility? A plain guide for founders

AI visibility is whether and how your product gets named, cited and described when someone asks an AI assistant for software to solve their problem. When a founder types "best tools for onboarding emails" into ChatGPT, Gemini, Perplexity or a Google AI Overview, the assistant returns a short list of named tools with a sentence of context for each. AI visibility is your presence in that answer: whether you appear at all, whether you are described accurately, and whether the assistant points to you as a credible option. It is distinct from a ranking. There is no page two to scroll to and no list of ten blue links. There is one synthesised answer, and you are either in it or you are not.

The jargon, in one place

Four terms get used for overlapping things, often interchangeably and often by people selling something. They are not synonyms, and the differences decide what work you actually do.

If the distinction between AEO and GEO feels thin, that is because it is. The split worth keeping runs elsewhere: AI visibility is the thing you measure, and the optimisation work is what you do about it. Measure first, because without a measurement you cannot tell whether your problem is that you are absent, that you are described wrongly, or that you are present on one engine and missing from another.

Why it matters, measured rather than asserted

We measured a set of customer questions about software on 22 and 23 August 2026. The results show how these companies appeared in that sample. They do not measure customer behaviour, visits or sales.

The sample covered 152 companies, with 30 customer questions prepared per company. ChatGPT and Gemini used an eight-question subset; Perplexity and Google AI Overviews used the full set, with errors excluded. Across the collected answers, the 18 established companies were named in 36.7% of answers about their questions. The 118 young companies were named in 6.4%, and 48 were not named in any of their collected answers. The median young company appeared in 1.4% of its answers. These are different summaries: the pooled answer rate is not the median company rate.

Naming and linking were separate events. Across the young and established cohorts, 1,068 answers named the company the question concerned. Of those, 653 cited its own website and 415 did not. That is a citation observation, not evidence that a reader clicked a link or that traffic was lost.

Across all 152 companies, the asked-about company’s own website appeared in 999 of 11,383 answers, or 8.78%. The same 999 appearances were 0.76% of the 130,694 recorded source slots. Here a source slot means one cited host in one answer; a host can recur in many answers. The two percentages have different denominators. Neither tells us how much publishing on your own site would improve visibility, or whether outreach to another site would work.

How AI assistants decide what to name

An assistant can use patterns learned during training and information retrieved for the question. An answer alone does not reveal the contribution of each.

What the model already knows

Training gives a model learned patterns about language, products and the world. Those patterns can affect an answer even when it contains no current source. Our study did not inspect training data or measure why a particular company was named, so it cannot attribute the difference between young and established companies to training alone.

What the assistant retrieves live

When an assistant uses search, it can retrieve information published after its training. A cited page is a useful place to inspect what the answer refers to. Open it and check the relevant passage: a citation does not establish that the page caused a recommendation or that the answer represents it accurately.

For a founder, the practical work is to make the product understandable and its claims checkable, then examine the answers to relevant questions. Clear documentation, original evidence and accurate descriptions elsewhere can each be worthwhile. Which one deserves attention depends on the gap you find.

The engines do not read the same internet

The collected source patterns differed by engine. In the full retained sample, community and user-generated hosts accounted for 2.6% of ChatGPT’s source slots, 4.5% of Gemini’s, 6.3% of Perplexity’s and 14.9% of Google AI Overviews’. ChatGPT and Gemini covered fewer questions, so these full-sample percentages are not a controlled comparison of engine behaviour.

The leading individual hosts in the full sample were techradar.com for ChatGPT, medium.com for Gemini, youtube.com for Google AI Overviews and reddit.com for Perplexity. Leading a list does not mean supplying most of an engine’s sources, and one host does not define the engine’s preferences for your market.

The retained analysis also compared top-ten host lists on the common eight-question subset per company. ChatGPT shared three hosts with Google AI Overviews, three with Perplexity and two with Gemini; Google AI Overviews and Perplexity shared six. This describes overlap in that sample, not the relative value of marketing on those sites.

How AI visibility differs from search rankings

AI visibility and search engine optimisation overlap but are not the same thing, and conflating them leads founders to do the wrong work. A search ranking is a position in an ordered list of links: your job is to earn the click, and the user then forms their own impression of you on your own page. An AI answer is a synthesis: the assistant has already formed and stated an impression of you on your behalf, in a sentence, before the user ever reaches your site. The unit of the contest changes from a position to a description.

Accuracy matters as well as presence. An answer that puts your product in the wrong category suggests different work from one that describes it correctly but gives no source. Clear product information and evidence help people assess your offer whichever route brought them there. This study did not compare the return from particular SEO or AI-visibility tactics.

How to tell if you have AI visibility

You cannot improve what you have not measured, and AI visibility is easy to measure crudely and worth measuring carefully. For the crude version, open two or three assistants, then ask them the questions a real customer would ask: name the problem you solve, ask for the best tools for it, ask for alternatives to your nearest competitor, and ask the assistant to describe your product directly. Read the answers as evidence, not as verdicts.

• Are you named at all when someone asks for tools in your category, or only when you name yourself first?
• When you are named, is the description accurate, or does the assistant put you in the wrong category or attach the wrong use case?
• When you are named, does the answer link to you, or does it link to a review site, a forum thread or a competitor's comparison page?
• Which competitors appear consistently, and what is being said about them that is not being said about you?
• Does the picture change from one engine to the next? Given how differently they source, it usually does.

Use the same questions when comparing assistants, and retain the date and the actual answers. Repeat checks before treating a pattern as stable. One absence may be worth investigating, but it is not a reliable estimate of how often you will appear.

How to start improving your AI visibility

Start with a relevant customer question and the answer it produces. The aim is to find a specific problem you can address, with a useful deliverable and a way to check what happened afterwards.

If the product is described incorrectly, check whether your own site clearly explains what it does, whom it serves and the evidence behind its claims. Correct an unclear or outdated page when that is the problem. A low share of citations to your own domain is not a reason to rule out this work.

If a relevant third-party page is incomplete or inaccurate, first read it and check that your product belongs there. A useful correction, comparison or contribution may be appropriate. If you lack evidence for the claim you want to make, the next move can be an experiment or a worked example. Creating something worth citing is different from asking to be included.

Choose the smallest useful next step. For example, an answer may repeat an outdated limit that still appears in your documentation. Correct the documentation, save the old and new wording, and later repeat the same question. If the answer changes, record that observation without claiming the edit alone caused it. Use visibility alongside useful visits, enquiries and sales when judging whether the work helped.

How we measured this

The retained run covered 152 companies: 118 young companies, 18 established companies and 16 AI-search products. Thirty customer questions were prepared per company. ChatGPT and Gemini used eight selected questions per company; Perplexity and Google AI Overviews used all thirty. Collection ran from 22 August 2026 at 20:44 UTC to 23 August at 14:43 UTC. The analysis retained 11,383 non-error answers and 130,694 source slots, counting a cited host once per answer. These are repeated observations across questions and companies, not 130,694 unique pages or independent samples.

The percentages describe this cohort, these questions and this collection window. Model and search behaviour can change, and automated mention detection can make mistakes. The pooled rates also weight companies and engines by their retained answer counts. They should not be read as the experience of a typical founder or as a probability for a new company. In the young and established cohorts, the own-site citation rates among answers that named the company were 69.8% and 51.4% respectively; this association does not establish why either group was cited.

We did not test an intervention in this run. It establishes neither the effect of publishing a page, getting a third-party mention or posting on a particular platform, nor any resulting traffic or revenue. The comparison is a starting point for investigation, not a prescription for where to spend your marketing time.

Where to go next

Check the questions that matter to your customers, inspect the answers and select a useful next move. Save what you changed and judge the result against the goal of that move.