GEO and AEO for B2B: measuring how often AI names you, and fixing the gaps
How to measure B2B AI visibility properly: which buyer questions to track, why one answer proves nothing, the metrics that matter across ChatGPT, Perplexity, Gemini and Google's AI features, and a fix loop that shows what worked.
AI visibility for B2B means one thing you can measure: when your buyers ask an AI assistant the questions that lead to a purchase, how often is your company named, in what position, next to whom, and based on which sources? Measure it on a fixed set of real buyer questions, across several engines, every day, because single answers vary. Then fix the gaps one question at a time and check whether the share of answers naming you moves.
Why one screenshot proves nothing
The most common AI visibility "report" is a screenshot: someone asked ChatGPT a question and you were or weren't named. It's close to useless as a measurement.
- Answers vary between runs. The same question asked twice can name different vendors in a different order, especially for comparison questions.
- Answers vary by engine. ChatGPT, Perplexity, Gemini and Google's AI features use different retrieval and different sources.
- Answers vary by phrasing. "Best close software" and "close software for NetSuite companies" can produce different lists.
So the unit of measurement isn't an answer. It's a share of answers: of all the times a question was asked, in each engine, over a period, how many named you.
Step 1: pick the questions that matter
Track buyer questions, not keywords. A good B2B question set has 10 to 50 questions across four types:
| Type | Example | Why it matters |
|---|---|---|
| Category | "Best accounts payable automation for mid-market" | Shortlist formation |
| Problem | "How do I cut month-end close from 10 days to 5" | Early research; who AI frames as the solution |
| Comparison | "Vendor A vs Vendor B for multi-entity" | Late evaluation; often decisive |
| Constraint | "AP automation that integrates with NetSuite and is SOC 2" | Narrow, high intent, easiest to win |
Sources for real questions: sales call notes, the questions prospects ask in demos, your ChatGPT Ads search terms, support tickets from new customers, and the "People also ask" style questions in your category. For more on question selection, see ChatGPT Ads buyer questions; the same questions usually work for both paid and organic.
Avoid tracking your own brand name as a question. "What is [your company]" tells you if AI knows you exist, not whether it recommends you.
Step 2: measure across engines, daily
Track at least these surfaces, because B2B buyers use all of them:
- ChatGPT
- Perplexity
- Gemini
- Google AI Overviews
- Google AI Mode
For Google's AI features, Google says "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary": the page must be indexed and eligible to show with a snippet. Google also notes that both features may use "query fan-out", issuing several related searches to build an answer, and that traffic from AI features is included in Search Console's Performance report under the Web search type (Google Search Central).
Daily runs smooth out the variation. Ten questions across five engines every day for 30 days is 1,500 answers a month, which is enough to see real movement in a share.
Step 3: the metrics that matter
| Metric | Definition | What it tells you |
|---|---|---|
| Mention rate | Share of answers that name you | The headline number |
| Position | Where you appear when named (first, second, later) | First mention gets the most attention |
| Competitor share | Share of answers naming each competitor | Who AI recommends instead of you |
| Cited sources | Which pages the engine linked or cited | Where AI gets its view of your category |
| Fan-out searches | The searches the engine ran before answering | The sub-questions your content needs to answer |
| Sentiment and framing | How you're described when named | "Enterprise-only", "expensive", "for small teams" |
Report mention rate per question and per engine, not one blended number. A 40% blended rate can hide 80% on Perplexity and 5% on Google AI Overviews, and those need different fixes.
Step 4: diagnose the gap
For each question where you're rarely named, look at three things in the answers:
- Who is named instead? If it's always the same two competitors, look at what they have that you don't: a comparison page, review site presence, a clear category statement.
- What's cited? If the engine cites third-party listicles, review sites or community threads, your own website isn't the main lever. You need to be on the pages AI trusts.
- What did it search for? Fan-out searches show the sub-questions. If ChatGPT searched "AP automation NetSuite integration" and you have no page that answers that clearly, that's the gap.
Step 5: fix one gap at a time
The fixes that move B2B mention rates are usually unglamorous:
- A direct answer at the top of the page that targets the question. Two or three sentences that answer it plainly and name your product with one concrete fact.
- Comparison pages. Honest "you vs. competitor" and "alternatives to competitor" pages. AI engines lean heavily on these for comparison questions.
- Constraint pages. "Works with NetSuite", "SOC 2 Type II", "for multi-entity groups": one clear page per constraint buyers ask about.
- FAQ sections with matching schema. Google says structured data should match the visible text, and that there's no special schema needed for AI features (Google Search Central). Schema helps machines parse; it doesn't replace content.
- Third-party presence. If answers cite review sites, industry lists or forums, get accurately listed there.
- Crawler access. Check that robots.txt doesn't block the AI crawlers you want to be read by, and that key pages render without heavy client-side scripts.
Do one fix per question, record the date it went live, and wait. Engines need time to recrawl and re-index; a week is a reasonable minimum before you read the effect.
Step 6: measure the effect, honestly
Compare the mention rate for the fixed question in the 7 to 14 days after the fix with the 7 to 14 days before. Keep two cautions in mind:
- Engines change on their own. Model and retrieval updates move rates for everyone. Track a few control questions you didn't touch.
- Small numbers swing. A move from 2 of 50 answers to 5 of 50 is promising, not proven. Look for direction across several questions.
Over a quarter, you learn which fix types work for your category. That's worth more than any single result.
Where paid fits
AI visibility and ChatGPT Ads are separate. OpenAI says "Ads do not influence ChatGPT's answers" (OpenAI Help). But they're complementary: while a fix takes weeks to show up in organic answers, a sponsored card below the answer can put you in front of the same buyer question today. See What ChatGPT Ads can't do for B2B.
Connect visibility to accounts
AI visibility only matters if it reaches buyers. Close the loop on your website:
- Visits from AI answers. ChatGPT, Perplexity and others pass referrers or UTM tags on some links. Track visits by AI source.
- Identify the companies. Company identification on those visits tells you which businesses arrived from AI answers.
- Use it in stages. A target account arriving from an AI answer is a stronger signal than a random visit; IntentBid lets you use "Visits from AI answers" as a stage condition. See ABM stages sales will trust.
How IntentBid helps
AI visibility in IntentBid asks your buyer questions every day in ChatGPT, Perplexity, Gemini, Google AI Overviews and Google AI Mode, and shows whether you're named, in which position, who's named instead, the cited sources, the full answers and the searches ChatGPT ran first. GEO audits check your pages for content and AI crawler access. On Growth and above, the autopilot proposes the next fix for each gap and the AI agent drafts it (answer block, FAQ with schema, comparison page outline), then measures the change once there are 7 days of data. The GEO agent writes the articles, pages and code fixes on request.
FAQ
What's the difference between GEO and AEO?
In practice they're used interchangeably. AEO (answer engine optimisation) emphasises being the answer; GEO (generative engine optimisation) emphasises being cited in generated answers. The work is the same.
How many questions should we track?
Start with 10 to 20 high-value questions. More questions dilute attention; you'll only fix a few per month anyway.
Can we pay to be named in ChatGPT's answers?
No. OpenAI says ads don't influence answers. Ads appear separately, below the response, and are labelled as sponsored.
Does schema markup get us into AI answers?
Not by itself. Google says no special schema is needed for its AI features. Schema helps engines parse pages that already answer the question well.
How long until fixes show up?
There's no fixed time. It depends on how quickly each engine recrawls the page and whether the fix is on your site or on third-party sources. Google notes that crawling "can take anywhere from several days to several months" (Google Search Central). Start reading a week after go-live and keep watching for a month.