Is a free AI visibility audit worth anything?

A free AI visibility audit is worth something specific and limited: it shows you what the engines currently say when somebody asks about your category, which most marketing teams have never systematically looked at. That is a real finding and it is usually enough to tell you whether this channel matters for your company. Treating it as a measurement of your visibility is where it goes wrong.

Every free audit in this category is lead generation. Ours is too, and saying so costs nothing, because the question that matters to a buyer is not whether a vendor hopes to sell you something. It is whether you keep anything of value if you do not buy.

What separates a useful free audit from a form is disclosure. Which prompts were asked, written out rather than summarised as a category. Which engines were checked, and how many times each prompt was run on each of them. And the raw answers, so you can read what the engine said instead of a vendor's account of it.

If an audit withholds all three and leads with a number out of 100, the number is unfalsifiable. You cannot tell whether a score of 34 means your pages are badly built or that the vendor asked three questions nobody searches. Scoring methods also differ between vendors, so a score from one firm cannot be compared with a score from another, which makes the headline figure the least portable thing in the report.

What can a free AI visibility audit actually tell you?

A free audit can establish four things reliably, and all four are worth knowing before you spend anything. Whether the engines name vendors at all for your category is the first. Some categories return a generic explainer instead of a shortlist, and if yours is one of them there is nothing to win yet, which we cover in does AEO work if nobody is searching for your category.

Who gets named instead of you is the second, and probably the most useful output of the whole exercise. A list of the three or four competitors the engines keep returning tells you who you are competing with in this channel, which is often not who you compete with in a pitch.

Anything the engines say about you that is wrong is the third. Outdated facts come up occasionally, and when they do they are cheap to fix and embarrassing to leave. Our page on why AI gets your business information wrong covers how that happens.

Last, whether your pages appear among the cited sources at all. That one distinction separates two different problems with different costs. A company absent from the sources has a retrieval problem. A company cited as a source but left out of the recommendation has a brand problem, and the two need different work, as our page on why ChatGPT recommends your competitor sets out.

None of those four requires many runs or a large prompt set, which is why they fit inside something a vendor can afford to give away. All four are observations rather than measurements, and that is exactly why they survive the sample size problem described below.

What can no AI visibility audit tell you, free or paid?

No audit can tell you your true AI visibility from a single reading, and no audit can predict how often you will be cited from the condition of your pages. Both limits come from published measurements rather than from modesty.

On the first limit, Schulte, Bleeker and Kaufmann tracked four AI search engines daily across four campaign verticals from 24 January to 20 March 2026, with a second dataset of identical prompts submitted repeatedly on the same day to separate run to run randomness from drift over time. Cited source sets overlapped between consecutive days by only 34% to 42%. Brand sets were steadier at 45% to 59%, still with wide variance. Their bootstrap convergence analysis found the standard error of a per brand detection rate drops below 0.10 at seven runs, and that source level coverage needs eight. Their phrasing is blunt: answers vary across runs, prompts and time, making one-off observations unreliable.

On the second limit, Chen and Koudas, across 158,099 answers and 468,037 citations, found that which pages the engine's own search retrieves explains about 32% of the variation in citations while page properties explain 2.8%. A page audit is scoring the 2.8%. That makes it a readiness check, which is useful, and not a forecast, which is what a score implies.

A third limit catches anyone promising one number. Benjamin Tannenbaum compared 589 citation observations across ChatGPT, Microsoft Copilot, Google and Perplexity from 15 commercial prompts and found that 84.9% of engine pairs shared no cited URL at all for the same prompt. Fifteen prompts on one day by an author who sells engine measured visibility is a small and interested sample, so read the figure as an upper bound rather than a point estimate. The direction is not in doubt: a single blended score is averaging four largely separate outcomes.

So the honest shape of any audit, including a paid one, is a diagnosis of condition plus a sampled observation of behaviour, each reported with its uncertainty. An audit that reports a share of answers with no interval around it is presenting a sample as a fact.

What should a free audit hand you before you take the sales call?

Ask for four artefacts, and ask before the call rather than during it. A vendor who will not send them is selling access to a conversation, not a diagnosis.

What to ask for	Why it matters	What a weak answer looks like
The exact prompts	Prompt choice decides the result. Questions nobody asks produce a flattering or alarming reading at will.	"We tested your category" with no list
Which engines, and how many runs each	Day to day source overlap runs at 34% to 42%, so a single run per engine is a sample of one.	A blended "AI visibility score" with no engine breakdown
The raw answers	The cited sources tell you whether your problem is retrieval or recommendation. A summary hides that.	Screenshots of a dashboard
What they would not fix	A vendor willing to name the parts outside their scope is describing a service rather than an outcome.	Everything is in scope and results are guaranteed

One more question is worth asking and almost nobody does: what would make you tell us not to buy this. Some categories are not ready, some sites have nothing indexable to work with, and some companies are better off fixing their classic search foundations first, which we cover in what to fix before you hire an AEO agency. A vendor with no answer to that question has no stopping rule, and a service with no stopping rule turns into a subscription.

If you want the version of this with the prompts and the raw answers on the table, our free visibility assessment is a 20 minute call on the questions a buyer asks in your category.

How do you tell a real free audit from a lead-capture form?

Tell a real free audit from a lead-capture form by what arrives before a human does. A real free audit sends you findings and then offers a call. A lead-capture form offers a call and produces the findings during it, which means the asset you were promised is the vendor's attention rather than information about your company.

The tells are reliable enough to list. An instant score from a URL alone is a page crawl with a label on it, not an audit of AI answers: nothing was asked of any engine, so nothing about your visibility was observed. A report that names no competitors has probably not run real buying questions, because buying questions return vendor names. And a report with no date and no engine list cannot be re-run, which means you cannot check it later or compare it with anything.

The commercial model is fair to state plainly. Free audits exist to find companies with a problem worth paying to fix, and paid audits are priced well above them because they add run counts, competitive benchmarking and a fix plan. Our own prices sit on the pricing page rather than in this article, because a figure copied into a guide goes stale the day it changes.

What you should not accept is a free audit that creates urgency out of noise. Given the 34% to 42% day to day overlap in cited sources that study found, a vendor can honestly report that you lost visibility this week and be describing the measurement rather than your company. Anybody presenting a single week as a trend either does not know that or is counting on you not to.

What does our own free assessment cover, and what does it not?

Our free assessment is a quick read, and the site says so in those words. As our pricing page describes it, we ask ChatGPT, Perplexity, Google AI Overviews and Claude the basic questions a buyer asks about you and your category, then walk you through the answers on a 20 minute call: where the engines name you, where they name competitors instead, and anything they say about you that is clearly out of date. We also say which of our two services fits, if either does.

What it is not is the full diagnosis. Our paid Accuracy Audit, for the separate problem of engines stating wrong facts about a company, agrees your facts with you first, checks each one on all four engines several times rather than once, checks the pages the engines actually cite, traces every wrong claim to its source and ends in a ranked fix list with replacement text. The difference between those two things is mostly run count and scope, which is the difference the sample size research above makes expensive to fake.

Two honest limits apply to ours as much as anyone's. A 20 minute call cannot carry seven or eight runs per prompt across four engines on a wide prompt set, so what you get is an observation of what the engines are saying, not a stable share of voice. And we will not know, from the assessment alone, how much of your gap is retrieval and how much is brand familiarity, although the cited sources in the answers usually indicate which way it leans.

Our page on what an AI visibility audit includes sets out what the longer version contains, and how many prompts you need to track AI visibility covers the sizing question behind the run counts.

When should you pay for an audit instead of taking the free one?

Pay for an audit when you need a number you can defend in a planning meeting, or when the decision in front of you is larger than the fee, which happens in three situations.

The first is a budget request. A free assessment produces a story, and a story is enough to start a conversation but not to justify an allocation. If you have to ask for money, you need a baseline with a stated method and a repeatable run count so that next quarter's reading means something against this one. Our page on getting budget approved covers what that paper needs to contain.

The second is a factual problem. If engines are stating something wrong about your company, the work is tracing each claim to the source that feeds it, and that is methodical rather than quick. Guessing at the source wastes a correction request on the wrong site.

The third is a decision with a long tail, such as a replatform, a rebrand or choosing between two agencies. A paid audit before any of those gives you a before reading, and without one you will never know what the change did.

Below those thresholds, the free version is the right purchase. Look at the answers, find out who gets named instead of you, check whether your pages appear in the sources at all, and come back when you have a decision that needs a measurement instead of a look.

What else do people ask about free AI visibility audits?

The five questions below come up when somebody has been offered a free audit and wants to know what they are agreeing to before they fill in the form.