What is a GEO audit? What it checks, what it costs, and what you get in 2026
- Definition: a GEO audit measures whether ChatGPT, Claude, Perplexity, Gemini and Google AI Overviews name your brand for your buyers' real questions, documents what the answers get wrong, and ends in a ranked fix plan.
- Cost in 2026: free automated scanners, $19 automated reports, hand-run audit + fix plan services at €149 to €690, agency one-off audits at $1,500 to $7,500, agency retainers at $1,500 to $8,500 a month (every price checked on 23 August 2026).
- What it catches: in our own audit work, an engine quoted a client's monthly prices as annual (about 25% too high) and public databases named four different CEOs for one company; automated scores marked both answers healthy.
- Turnaround: a scanner returns a score in about two minutes; a hand-run audit takes 3 to 7 business days.
- Do it yourself first: ten buyer questions across five engines in an afternoon tells you whether you appear; it cannot tell you why, or what to change first.
Someone on your team asks ChatGPT which providers to shortlist in your category, and your company is missing from the answer, or present with a price you retired last year. A GEO audit is the structured check that tells you how often AI engines such as ChatGPT, Claude, Perplexity, Gemini and Google AI Overviews name your brand for your buyers' real questions, what those answers get wrong about you, and what to change first. The part worth knowing before you buy one is that the label covers everything from a two-minute automated score to a multi-week agency engagement, and the deliverables differ more than the prices suggest.
The rest of this page covers what a GEO audit is, the six things a good one checks, and what the market charges in 2026, with every price checked on 23 August 2026. It then shows how to run a rough version yourself in an afternoon, and how to tell a report that changes something from a score that changes nothing.
What is a GEO audit?
A GEO audit (GEO stands for generative engine optimization: getting brands named in AI answers) is a structured review of how AI engines answer your buyers' real questions: which brands get named, in what order, with what reasoning, and what each answer claims about you. The output is a share-of-voice picture across engines, a list of documented errors traced to their sources, and a fix plan ranked by impact.
The audit exists because buying research has moved into chat windows. When a buyer asks an AI engine which provider to use, the reply contains three or four names with a reason attached to each. There is no page two behind it and no analytics report listing the answers you lost. Search Console shows you a lost ranking, while nothing in your stack shows you a lost recommendation, and a GEO audit is the instrument that makes that conversation visible and measurable.
One naming note helps you compare offers. GEO audit, AI visibility audit and AEO (answer engine optimization) audit are sold as near-synonyms, and in 2026 all three labels are in use. Some providers draw a line, with an AEO audit checking how well your pages are structured for answer extraction and a GEO audit checking whether the brand actually gets cited. Most offers cover both, so judge an offer by the checks and deliverables below rather than by the label on the invoice. If GEO itself is new to you, our plain-English guide to generative engine optimization covers the basics.
What does a good GEO audit check?
A good GEO audit checks six things: share of voice per engine, the accuracy of what the answers claim, how the answers frame you, whether AI crawlers can read your site, what your off-site record says, and which sources each answer was built from. An offer that skips several of these is a scan rather than an audit. The table shows what each check catches, with examples from our own audit work (clients anonymized).
| Check | What it measures | What it catches (real examples) |
|---|---|---|
| Share of voice, engine by engine | For each buyer question and each engine: who is named, at what rank, how often | In a 25-answer panel for an ecommerce software client, the client appeared in 0 of 25 answers while the category leader appeared in 20 and the runner-up in 18. A single blended score would have hidden which questions were winnable. |
| Accuracy of claims | Every price, plan, feature and company fact stated about you, checked against reality | ChatGPT quoted one client's monthly prices as annual, so every quote read about 25% too high. The answer looked healthy to any automated scanner. |
| Framing and sentiment | The adjectives, hedges and comparisons that travel with your name | "Affordable but dated" and "premium and polished" win different deals. The audit records the wording of every answer so the pattern is visible before a buyer repeats it to you. |
| Technical access | Whether AI crawlers can fetch and read your pages: robots.txt, crawler response codes, content that survives without JavaScript, structured data, sitemap | Our own host's CDN returned HTTP 429 to GPTBot and Perplexity's crawlers on several pages, and homepage stat counters rendered as "0" in raw HTML, so a crawler quoted zeros. Neither was visible from a browser. |
| Off-site record | Directories, review platforms, company databases, marketplaces and community threads: what each says about you and whether they agree | Public company databases named four different CEOs for one client. In the 25-answer panel, the client was absent from all 18 third-party surfaces checked. |
| Source tracing and per-engine differences | Which pages each answer was built from, and how the five engines differ on the same question | Four of five cost answers quoted competitor prices word for word from public pricing pages. One engine still recommended a vendor whose domain no longer resolves, because directories and old roundups carried the name. |
The sixth check is where an audit earns its fee. Knowing an answer is wrong is useful. Knowing which directory profile, retired page or stale roundup fed it turns the fix into a ten-minute task instead of a guessing game, because the engines keep reading that source until someone changes it.
What does a GEO audit cost in 2026?
A GEO audit costs anywhere from nothing, for an automated scanner, to $7,500 for a full agency audit, and agency retainers run $1,500 to $8,500 a month on top of that. Hand-run audit + fix plan services sit between those poles; our own run €149 for a five-question Snapshot and €690 for a 30-question Full Audit. Every figure below was checked against the provider's public page on 23 August 2026, and the endpoints are named so you can check them yourself.
| Tier | Price (checked 23 Aug 2026) | What you get | Named endpoints | Best for |
|---|---|---|---|---|
| Free automated scanners | €0 | An instant visibility score from a small, generic prompt set | GoVISIBLE free audit, HubSpot AEO Grader, GeoAnalyzer free score | A first curiosity check |
| Automated paid reports | About $19 one-time | A PDF score with generic recommendations and a template roadmap | GeoAnalyzer Full Report, $19 | Checking whether a score moved |
| Monitoring tools | $29 to $489 per month | Scheduled prompt runs and a dashboard; extra engines often cost extra | Otterly.AI Lite $29 (15 prompts, 4 engines; Claude add-on $29 to $439), Standard $189, Premium $489; GoVISIBLE GEO Flex $79 (50 prompts, 5 engines) | Watching a number after fixes ship |
| Hand-run audit + fix plan | €149 to €690 one-off | Your real buyer questions run by hand across five engines, every claim verified, errors traced to source, a ranked fix plan | MentionShare Snapshot €149 (5 questions, about 3 days) and Full Audit €690 (30 questions, 5 to 7 days, 7 monthly re-measures included). Disclosure: that is us. | Brands ready to act on the findings |
| Agency one-off audits | $1,500 to $7,500 | 10 to 30 prompts across 2 to 5 platforms, usually with workshops and a deliverable package | Published agency ranges: Demand Local (April 2026) lists $1,500 to $3,000 for 10 prompts on two platforms and $5,000 to $7,500 for 30 prompts on five; Mentionable (May 2026) lists one-shot audits at $2,000 to $4,500 | Teams that want implementation bundled |
| Agency retainers | $1,500 to $8,500 per month | Continuous monitoring of 20 to 200 prompts plus ongoing optimization work; enterprise programs start above $10,000 | Mentionable's May 2026 pricing survey | Enterprises with several markets or languages |
The spread comes down to manual hours and scope. Reading and verifying hundreds of AI answers takes days of a person's time, and each added engine, competitor, market or language multiplies the panel. Tools spread their software cost across thousands of subscribers and charge per engine instead (Otterly prices Claude as an add-on at $29 to $439 a month depending on tier), while agencies bundle workshops and implementation into the fee. Otterly's and GoVISIBLE's prices were unchanged between our 14 August and 23 August checks, and they are still worth re-checking before you buy, because the category is young and plans move. Sources for the agency ranges: Demand Local's GEO audit walkthrough and Mentionable's GEO agency pricing guide.
How do you run a rough GEO audit yourself?
You can run a rough GEO audit in an afternoon with a spreadsheet and the five engines' free tiers, and it will tell you whether you appear at all and roughly how the answers describe you. The steps below are the same first stage we run by hand on every engagement. What they cannot give you is source tracing and a prioritized fix plan, and the note after the steps says exactly where that line falls.
- Write ten questions your buyers actually ask, in their words. Pull them from sales calls, support tickets and the search queries that already bring you leads, and make at least three of them comparison questions ("best X for Y", "X alternatives"), because those decide shortlists. Done when you have ten questions and none of them contains your brand name.
- Run every question through ChatGPT, Claude, Perplexity, Gemini and Google AI Overviews in a fresh or logged-out session, so personalization does not color the answers. Done when you have 50 answers saved as text, each with its date.
- Record, for each answer, every brand named, its position and the reason given. A sheet with one row per answer and columns for ranks one to five is enough. Done when you can read your mention rate per engine straight off the sheet.
- Check every claim about you against your real facts: prices, plans, features, company details, people. Paste the exact wording of each wrong claim next to the correct fact. Done when you have a list of errors with the offending sentence beside each one.
- List the sources each engine shows (Perplexity cites by default; Google AI Overviews and ChatGPT's search mode usually show links) and open every one. Done when you know which pages fed the answers, and which of those pages are yours.
- Run two free scanners on your brand and your closest competitor (GoVISIBLE's free audit and HubSpot's AEO Grader both run without a card) and keep the scores as a second data point. Done when three numbers are written down: yours, theirs, and the gap.
The DIY version stops at presence and obvious errors. It misses the sources behind answers that cite nothing (Claude and Gemini vary in what they show) and the third-party profiles that feed answers without being cited. It also misses the variance between runs, since the same question produces different answers on different days and a single run carries error bars. And it misses the prioritization of twenty problems into the three to fix this week, with the corrected text written. Those four gaps are where a paid audit spends its hours, and our step-by-step guide to getting cited by ChatGPT and Perplexity covers what to do with whatever the afternoon turns up.
Automated scanner or human-run audit: which do you need?
A scanner measures how often you appear; a person also reads what the answers say, notices when they are wrong, and traces where the wrong fact came from, and wrongness is where deals quietly die. Run a free scan whenever you like and treat it as a thermometer. Bring in a human-run audit when you want to know what the answers actually tell your buyers and what to change, in what order.
Two findings from our own work this summer show the difference. An engine answered a pricing question by reading a client's monthly prices as annual, so every quote a buyer heard was about 25% too high. Public company databases named four different people as the same company's CEO, and the engines repeated whichever record they happened to pull. No automated score surfaces either problem, because both answers looked perfectly healthy and were simply false.
Our audits are built for exactly that gap. MentionShare is an AI-visibility (GEO) audit service for B2B and B2C brands. Laurynas Leskauskas runs your buyer questions by hand through all five engines, reads every answer, verifies every claim against your real facts and traces each error to its source. He then writes the fix plan, with the page openings, correction emails and disclosed replies already drafted. The Snapshot (€149, five questions, about three business days) is the way to test the method on your own category. The Full Audit (€690, 30 questions, 5 to 7 business days) adds the full leaderboard with mention rate, rank and sentiment, every error traced, 15 or more prioritized fixes and seven monthly re-measures of the same questions. Both carry a 14-day money-back guarantee, you can read a complete sample report before paying anything, and our methodology write-up shows the hand-run method step by step.
One honest limit applies to every audit, ours included: it cannot guarantee rankings or mentions. Engines change and model memory moves on its own timetable, so the audit's job is to make the fastest-moving layer, live retrieval, work in your favor with fixes whose effect you can re-measure.
What should a GEO audit hand you when it is done?
Judge any GEO audit by what lands in your hands. You should receive four things: a dated, per-engine leaderboard against named competitors; every error documented with its evidence and likely source; a fix plan ranked by impact, with an owner and an effort estimate per fix; and a commitment to re-test the same questions after the fixes ship. A report missing two of those four is a scan with a cover page.
- A dated, per-engine leaderboard. You against named competitors, question by question, so the baseline can be re-tested later with the same questions.
- Every error documented with evidence. The exact claim, where the engine most likely picked it up, and the correct fact, so your team fixes the source instead of guessing.
- A fix plan ranked by impact. Each fix with one owner (marketing, developer, founder), an effort estimate, the text or email ready to use, and a "done when" check, ordered so the quick wins come first.
- A re-test commitment. The same questions run again after your fixes, because a baseline without a follow-up is just a photograph.
Three red flags are worth walking away from: a single score with no underlying answers to read, advice so generic it could ship to any company in any industry, and a "tested" claim with no dated answers shown. Our comparison of the best AI visibility audit services in 2026 applies exactly this test to eight providers, ourselves included.
How often should you repeat a GEO audit?
Plan on a re-measure within a month or two of shipping your fixes, then a steady rhythm after that. Models update, retrieval sources change, and competitors publish. Fixes on pages the engines fetch live can show up in answers within days to weeks, while corrections to third-party databases take longer to propagate. The first re-measure therefore tells you which fixes landed and which sources still need chasing.
Monthly monitoring makes sense while you are actively climbing, and quarterly checks are enough once your category is stable and spot-checks keep coming back accurate. Auditing once and framing the certificate is the one pattern that fails, because the answers keep moving whether or not anyone is watching them. A Full Audit from us includes seven monthly re-measures for this reason, and a monitoring tool at $29 to $79 a month covers the same need for a brand that has already fixed its findings.
The short version: what a GEO audit is worth
A GEO audit is the one document that shows you the buying conversation your analytics cannot see: who the engines name for your buyers' questions, what they get wrong about you, and which sources to fix first. If the budget is zero, run the afternoon version above and two free scanners, and you will know whether you have a problem. When you are ready to act on it, a hand-run audit + fix plan turns that problem into a ranked list with the text already written. For most brands, that first baseline pays for itself with the first wrong price it catches. How GEO differs from SEO explains why none of this shows up in the tools you already run.