By Lucan Ashvere. Updated October 2026. A free method you can run yourself, no tools required.
More B2B buyers now start a software search by asking an AI assistant "what's the best tool for X?". If your product isn't named in that answer, you may not make the shortlist, and your analytics won't tell you, because the buyer never clicked anything. Here is a simple, repeatable way to find out where you stand in ChatGPT, Perplexity, Gemini and Google AI Overviews, using only a spreadsheet.
Step 1: Write 10 buyer prompts (10 minutes)
Use the words a buyer would type, not your marketing language. Cover three stages of the buying journey:
- Category discovery (4 prompts): "best <category> software for <company type>", "top <category> tools 2026", "<category> tool for <specific use case>".
- Alternatives (3 prompts): "<biggest competitor> alternatives", "cheaper alternative to <competitor>", "<competitor> vs <other competitor>".
- Validation (3 prompts): "is <your product> good for <use case>", "<your product> pricing", "<your product> vs <competitor>".
The first two groups are the ones that matter most: they show whether you're recommended when the buyer doesn't already know your name.
Step 2: Run each prompt in a clean session (20 minutes)
- Use a private/incognito window or a logged-out session where possible, and turn off memory or personalization. Your own account has probably seen your product name many times.
- Run every prompt in ChatGPT (with search on), Perplexity, Gemini and a Google search that shows an AI Overview.
- Run the discovery prompts at least twice. AI answers vary from run to run, so a single answer is an anecdote, not a measurement.
Step 3: Record four things per answer
| Column | What to write |
|---|---|
| Named? | Yes/No: does the answer name your product? |
| Position | 1st, 2nd, 3rd… in the list, or "mentioned in passing" |
| Competitors named | Every other product in the answer |
| Sources cited | The URLs the engine links to (Perplexity, ChatGPT search and AI Overviews usually show them) |
Also note anything the answer gets wrong about you: outdated pricing, a discontinued plan, a feature you don't have, or being described as a different category. Wrong facts are often easier to fix than absence.
Step 4: Score it (5 minutes)
Share of voice = answers that name you ÷ total answers, for the discovery and alternatives prompts only. Do the same for your top three competitors. A typical first result looks something like "we're named in 4 of 28 answers, our main competitor in 22." That gap is the problem to solve.
Then sort the Sources cited column by frequency. In most B2B categories, a small set of pages (review sites, "best X tools" listicles, comparison pages, Reddit threads, documentation) keeps getting cited. Those pages are what the engines lean on. If you aren't on them, that is usually the first fix.
Step 5: Check the technical basics (5 minutes)
- robots.txt: open
yourdomain.com/robots.txtand check you aren't blocking the crawlers AI search uses, such asOAI-SearchBot(ChatGPT search) andPerplexityBot. Blocking training crawlers likeGPTBotis a separate decision; read each vendor's documentation before changing anything. - Rendering: view the page source of your pricing and feature pages. If the key text only appears after JavaScript runs, some crawlers may not see it.
- Consistency: compare how your homepage, G2/Capterra profile, LinkedIn and Crunchbase describe what you do. Engines combine these sources, and conflicting descriptions lead to vague or wrong answers.
- llms.txt: it's cheap to add, but treat it as optional. Support among the major engines is limited and not well documented, so don't expect it to fix visibility on its own.
What usually moves the needle
- Getting onto the pages engines already cite: the listicles, review sites and comparison articles in your sources column.
- Honest comparison and alternatives pages on your own site that answer "X vs Y" and "X alternatives" directly, with specifics such as pricing, limits and integrations.
- Clear, consistent category language: say plainly what you are and who you're for, the same way everywhere.
- Fixing wrong facts at the source (old pricing pages, outdated review profiles).
Caveat: nobody controls what an AI engine says, and answers change between runs, users and weeks. Re-run the same prompt set monthly and track the trend rather than any single answer.
Want this done for you?
The method above is a cut-down version of what we do in a paid audit: 30 to 100 prompts, up to 6 engines, multiple runs each, a cited-source map, and a ranked fix plan, delivered in writing with no calls. See packages and pricing (from $690), or email lucan@business-insights.ai.