What Is AI Visibility? How to Measure It Without Buying a Tool

Published: September 7, 2026 | Last updated: September 7, 2026 | 13 min read

By: Graeme Whiles

You typed your own brand name into ChatGPT last month, read the answer, and felt something between relief and mild panic.

Most people I speak to have done exactly that.

Then they did it again a week later and got a different answer, which is where the mild panic usually wins.

However, the thing that follows is almost always the wrong move: buying an AI visibility tool before deciding what would count as a good result.

So here is the short answer before the long one.

AI visibility is the extent to which your brand appears in AI-generated answers, across platforms like Google AI Overviews, ChatGPT, Perplexity, Gemini and Microsoft Copilot. It covers three things: mentions, where a model names you with no link; citations, where it links to your page; and recommendations, where the answer puts you forward as the option to choose.

I understand the instinct. When a channel appears this fast, paying someone for a number feels like control.

The trouble is that half those numbers are estimates of estimates, and the two most credible sources of AI visibility data are now free and sitting in accounts you already own.

In this guide, I'll give you a definition of AI visibility you can actually work with, show you what the search results for the term look like right now, and explain which metrics survive scrutiny.

Then I'll walk through how to measure your current AI visibility using Google and Bing's own reporting, before you spend a penny on a platform.

Short on time? Here are the key takeaways

• AI visibility is presence, not position: it is whether your brand appears in AI-generated answers across Google AI Overviews, ChatGPT, Perplexity, Gemini and Microsoft Copilot, covering mentions without a link, citations with one, and recommendations that name you as the choice. There is no result number seven to settle for.
• You no longer have to buy a tool for the baseline: Google rolled its generative AI performance report out worldwide on 31 August 2026, and Bing has had an AI Performance report since February.
• Both first-party reports stop at impressions and citations: neither gives you clicks, which is exactly the gap AI visibility platforms are selling into.
• The volatility is real but mostly a prompt problem: Peec AI's 37,804-response study found brand mentions hold steady as long as your prompts keep the same intent.

What is AI visibility?

'AI visibility' means brand presence inside the answer itself, rather than in a list of blue links underneath it, across every AI platform people now use in place of a results page.

That's it. No mysticism.

Where it gets useful is the word 'appears', because it covers three quite different things and most marketing teams collapse them into one number.

• A mention: an AI model names your brand in its answer with no link attached. It builds recognition, it feeds branded search volume, and it is invisible in every analytics package you own.
• A citation: the AI platform links to your page as a source, which is the only one of the three that can send you a click.
• A recommendation: the answer actively puts you forward as the option to choose, which is the version with revenue attached to it.

Being cited is not the same as being recommended, and I've audited plenty of sites that got the first without ever getting the second. The term for the practice of improving all this is answer engine optimisation.

I use AEO, and I've written separately on what AEO is and how answer engine optimisation works in practice.

AI visibility is the outcome. AEO is the work.

How AI visibility differs from traditional search

Traditional search engines trained all of us to think in positions.

You were fourth, you wanted to be first, and there was always a position seven to fall back to while you worked on it.

AI answers do not have a position seven.

Google's AI Overviews, ChatGPT, Perplexity, Microsoft Copilot, and Gemini are all AI answer engines, and each returns a single, composed answer assembled from sources the model selects.

Your brand is in that answer or it is absent, and absent looks identical to not existing.

That single structural difference is what makes AI visibility a different measurement problem rather than traditional search rankings with a new coat of paint.

Three consequences follow, and they're the ones that catch marketing teams out.

• Keyword rankings stop being the unit: you are measuring share of voice across a set of prompts, not a position for a keyword. The question changes from "where do I rank" to "how often do I show up, and against whom".
• The click may never happen: user behaviour moved before the reporting did. SparkToro's June 2026 analysis of Similarweb clickstream data put zero-click searches at 68.01% of US Google searches between January and April 2026, against 60.45% in 2024. On the searches where an AI Overview appears, that study found click-through rates fall by close to 60%.
• Traditional SEO tools measure the wrong surface: a rank tracker watching position four tells you nothing about whether the AI Overview above it named a competitor.

Worth being clear on one thing, because the AEO industry has an interest in blurring it.

None of this makes traditional search engine optimisation obsolete.

The signals AI engines lean on are largely the signals Google already rewarded, which is why a site with broken structured data and thin pages does not have an AI visibility problem so much as an SEO problem wearing a newer costume.

AI visibility tracking: the measurement problem nobody leads with

"Our AI visibility reports feel made up. Am I the only one?" on r/DigitalMarketing. "AI visibility tools are a waste of money" on LinkedIn. 

Those people are not being cynical. They have watched the same prompt return different brands on Tuesday and Thursday and drawn the obvious conclusion.

So is AI visibility tracking measuring anything real?

Mostly yes, and the best evidence I've seen is a paper by Jan Ehrlinspiel, Malte Landwehr and Tomek Rudzki at Peec AI, published on 15 June 2026. It analysed 37,804 AI responses across 1,754 prompts, spanning ChatGPT, Gemini, Perplexity, Google AI Mode and Google AI Overviews, in 18 sub-verticals.

Its findings are the ones that should shape how you set your tracking up.

Brand visibility stayed stable as long as reworded prompts kept a cosine similarity (a measure of how close two phrasings are in meaning) above roughly 0.50 to 0.60. Only when prompts drifted below that, to 0.35 to 0.39, did mentions fall sharply, dropping around half against a 4.9% baseline mention rate.

Prompt format moved things too. Asking for a ranked list produced up to 20% more brand mentions than an open question, and keyword-style queries produced up to 25% more than conversational ones.

The randomness people are seeing is real, and most of it comes from tracking a scattered set of prompts that don't share an intent, then averaging the results into a single score. Prompts with the same intent and the same semantic shape return broadly the same brands at broadly the same rate.

Which means the fix is prompt design, not abandoning measurement. Track fewer prompts, keep each one anchored to a single buying intent, and stop reading day-to-day movement as signal.

How to measure AI visibility for free

On 31 August 2026, Google finished rolling out its generative AI performance report in Google Search Console to every site worldwide, after starting with UK site owners in June.

It reports impressions for AI Overviews and AI Mode, broken down by page, country, date and device.

That is genuinely useful, and it is also genuinely limited.

There are no clicks in it. No click-through rate, no average position, no query-level data. You can see that Google showed a link to your site inside an AI feature, and nothing about what happened next.

However, Google's own help documentation is upfront about this, and the UK's Competition and Markets Authority has since required Google to provide click-through data and page-level controls by March 2027.

Microsoft got there first, in February 2026, with the AI Performance report in Bing Webmaster Tools. It covers citations in Microsoft Copilot, Bing's AI summaries and selected partner integrations, showing total citations, average cited pages, the grounding queries the AI used to retrieve your content, and citation counts for individual URLs.

Then in June 2026, Microsoft added four things worth having: intent classification, topic clustering, period-on-period comparison, and Citation Share, which is the percentage of all citations on a given query that went to your site.

Citation Share is the closest thing to free share of voice anyone currently offers, and Microsoft is upfront about what it isn't. It does not name the competitor domains taking the rest, it is not traffic share, and it is not a quality score.

No click data there either.

So here is where the AI visibility platforms actually earn their money, and where they don't.

• What you should not pay for: impressions in Google's AI features, and citation counts and citation share in Copilot. Both are now first-party, free, and more accurate than any scraped estimate of them.
• What is still worth paying for (I still go back and forth on this though): coverage of ChatGPT, Perplexity, Claude and Meta AI, which publish nothing, plus named competitor share of voice and sentiment across a prompt set. Bing will tell you your citation share but not who took the rest, and that is a real gap.
• What nobody can sell you yet: reliable click and revenue attribution from AI answers. If a vendor claims it, ask which of the two platforms that hold the data is supplying it to them.

Most sites I audit are paying a subscription for the first bullet while ignoring the second.

The AI visibility metrics that survive scrutiny

Every AI visibility platform ships a proprietary score.

However, I would not build a reporting line on one, because you cannot audit the formula and it cannot be compared to anything.

So here is how to track AI visibility without one. These are the key metrics I actually use, in the order I'd add them.

1. Presence rate: the percentage of your tracked prompts where the brand appears at all, per platform. This is your AI presence in its plainest form. One number per AI platform, never blended into a single figure, because ChatGPT and Google AI Overviews reward quite different things.
2. Share of voice: how often you appear against the named competitors appearing in the same answers. It is the closest thing to competitive positioning you can measure in AI search, and it is the metric that tells you whether you're gaining or the category is simply growing.
3. Citation rate versus mention rate: the split between AI citations that link you and answers that merely name you. A high mention rate with a low citation rate is a real result, and it shows up as branded search volume rather than referral traffic.
4. Accuracy and sentiment: whether what the AI systems say about you is correct and how it frames you. A confident, wrong description of your pricing is worse than absence, and it is fixable.
5. AI bot traffic in your server logs: which AI crawlers are fetching which pages, and how often. This is the earliest available signal and the one nobody bills you for.
6. Branded search volume: the downstream proof. Mentions without links still create demand, and it lands in Search Console as people searching your name.
7. Assisted conversions in Google Analytics: imperfect, worth watching anyway. Buyers discover in an AI answer, then search your brand or type the URL, so your analytics credits organic or direct for a sale the assistant created.

Notice what is missing from that list.

There is no "AI visibility score".

The brands that get this right aren't the ones with the prettiest dashboard; they're the ones who can name the ten prompts their revenue depends on and tell you what happened on each of them last month.

What AI visibility looks like on a brand I built from zero

I'd rather show you a site I own than a client's, so here is Three Putt Golf Clothing, the golf clothing brand I built from a blank domain as a live proof of concept.

Start with what the AI platforms are doing, because this is the part no analytics package will show you.

Across the AI engines, Three Putt currently holds 3,474 citations spread over 709 pages. Google AI Mode accounts for 757 of them, Gemini 627, Google AI Overviews 572, Perplexity 531, Microsoft Copilot 493 and ChatGPT 276.

Now look at the shape rather than the totals, because that is where the useful part is.

Perplexity has taken 531 citations from only 47 pages, which is 11 citations per cited page. Google AI Mode has taken 757 from 411 pages, which is under two. Gemini sits at seven, Copilot at five, ChatGPT at three.

The same store is a completely different shape on every platform.

Perplexity has decided a handful of pages answer almost everything and keeps returning to them. Google spreads itself thinly across the whole catalogue. If I collapsed that into one blended "AI visibility score" I would lose the only finding on the page worth acting on, which is that Perplexity rewards depth on a few pages and Google rewards coverage across many.

That is the whole argument against a single number in one dataset.

Now the first-party half, from Search Console.

Between April and August 2026, the store took 1,698,207 impressions and 24,528 clicks at an average position of 6.7.

That is a click-through rate of 1.44%.

Sitting around position six or seven used to return something in the region of three to five per cent. Three Putt is earning roughly a third of the clicks its positions would once have paid out, and it is doing that while being cited 3,474 times in AI answers.

Both of those things are true at once, and neither number explains the other on its own.

That gap is what AI visibility looks like in the data before anyone sells you a tool to describe it.

You can check out the whole case study here.

How to improve AI visibility

Improving is a longer subject than defining, so here is the short version and where the detail lives.

• Say what you are, identically, everywhere: one canonical sentence on your homepage, about page, Organization schema and llms.txt file. AI models quote consistency, and I've watched ChatGPT return a client's own sentence back to me almost word for word.
• Publish declarative, quotable content: verdict first, evidence second, in a predictable structure. My guide to how to rank in ChatGPT covers the format.
• Fix the structured data before the clever bits: one claimant per fact, no conflicting schema blocks, per my guide to structured data for SEO.
• Earn third-party corroboration: most of what an AI platform says about your brand comes from pages you don't own. You cannot astroturf this at any scale worth having.
• Let the retrieval crawlers in: blocking OAI-SearchBot or PerplexityBot in robots.txt opts you out of the answer entirely, and AI Overview optimisation covers the crawler list properly.

Then measure it monthly rather than daily.

Should you be tracking AI visibility at all?

Yes, but this depends on what stage your business is at.

If your key pages run default metadata, your schema disagrees with itself, and you have never opened Search Console, AI visibility optimisation is not your next job.

The AI engines read the same site Google does, and nothing about the AI era changed what a well-built page looks like.

Fix the foundations first, and treat AI visibility as the layer that goes on top of a site that already works rather than a way around building one.

Final Thoughts

AI visibility is not a new discipline so much as an old question asked of a new surface: when somebody is deciding, does your brand come up?

The honest position is that the answer is now partly measurable, free, and from the source.

Google will tell you your impressions. Bing will tell you your citations.

Neither will tell you what it was worth, and anyone claiming otherwise is selling you a model of the thing rather than the thing.

So define your ten prompts. Pull the two free reports. Write down where you stand this month, and stop refreshing the dashboard in between.

Then do the unglamorous work underneath it, because AI platforms cite the sites that were already worth citing.

If you'd like to know exactly where your brand stands across ChatGPT, Google AI Overviews, Perplexity and Copilot, that's what my AI Visibility Audit covers.

Crawler access, structured data, canonical language, and where the AI engines currently place you against your competitors.

You can also run my free AEO readiness score in two minutes to see which foundations are missing first.

Get in touch, or email me directly.

Author Bio

Graeme Whiles is an independent SEO and AEO consultant at GWContent. He has worked with enterprise and SaaS brands, including Originality.ai, Connecteam, 6sense, Practice Better, and Peppr, growing organic traffic and AI search visibility across some of the most competitive categories in B2B. He also built Three Putt Golf Clothing from a blank domain as a live proof of concept for his methodology.

AI Visibility Frequently Asked Questions

What is AI visibility?

AI visibility is how often and how prominently your brand appears in AI-generated answers across platforms such as Google AI Overviews, ChatGPT, Perplexity, Gemini and Microsoft Copilot. It covers three distinct things: mentions where you're named without a link, citations where the AI platform links to your page, and recommendations where the answer actively puts you forward.

How is AI visibility different from traditional SEO?

Traditional SEO measures position in a ranked list of blue links, where there is always a lower position to occupy. AI answers have no positions. Your brand appears, or it doesn't, so the unit of measurement becomes presence and share of voice across a set of prompts rather than keyword rankings. The underlying signals overlap heavily, which is why traditional search engine optimisation still matters.

How do I measure AI visibility for free?

Start with the two first-party sources. Google Search Console's generative AI performance report, rolled out worldwide on 31 August 2026, shows impressions in AI Overviews and AI Mode. Bing Webmaster Tools' AI Performance report, live since February 2026, shows citation counts in Copilot and, since June 2026, your citation share on a query. Neither gives you clicks. Add your server logs for AI bot traffic and Search Console for branded search volume, and you have a baseline before spending anything.

Are AI visibility tools worth it?

For ChatGPT, Perplexity, Claude and Meta AI, yes, because those platforms publish no first-party data and there is no other way to see them. For Google and Bing impressions, no, because you can now get that free and more accurately from the source. Treat any single proprietary "AI visibility score" as a marketing artefact rather than a metric, since you cannot audit the formula.

What is a good AI visibility score?

There isn't one, and that is the honest answer. Every platform calculates its own AI visibility score differently, weighting mentions, citations and position in ways you cannot inspect, so a 72 on one tool and a 41 on another can describe the same reality. Judge yourself against your own presence rate last month, and against the competitors appearing in the same answers, rather than against a number you cannot audit.

Why do AI visibility results change every time I check?

Partly because AI models are non-deterministic, and mostly because of prompt design. Peec AI's June 2026 study of 37,804 responses found brand mentions stay stable when reworded prompts keep the same intent, and only collapse when the phrasing drifts far enough to change what is being asked. Track a small set of intent-anchored prompts and read monthly trends, not daily movement.