10/05/2026
A buyer opens ChatGPT and types, “Who should I hire for [your service] in [city]?” You have no idea what comes back. Most owners don’t. An AI visibility audit changes that. You write a fixed set of buyer questions, run them the same way monthly, and log three things: whether you’re mentioned, whether you’re linked, and which sources get cited.
People call this “LLM SEO” or “ChatGPT SEO.” The label doesn’t matter. You need a spreadsheet and some careful hours this week. A checklist sits at the end.
What an AI visibility audit measures
An AI visibility audit checks what AI answers say about your business and whom they point to. It sits next to your regular SEO work.
Why it is worth an hour a month
Buyers are already using these tools to build shortlists. A G2 survey of 1,076 B2B software buyers found 51% start research with an AI chatbot more often than with Google, and 69% picked a different vendor than planned after a chatbot recommendation (G2, 2026). That’s software buyers; your market may differ. A separate G2 report covering 1,038 decision makers put review sites (38%) and AI chatbots (37%) as the top two shortlist influences (Moore, 2026).
The three things you log
Every answer gets three checks. Were you mentioned? Were you linked? Which sources did it cite? You need your own log because Google doesn’t give you one. Traffic from AI features is counted in the “Web” report in Google Search Console, with no separate AI tab (Google Search Central, 2025). So the spreadsheet is the record.
Step 1: Write your 25 prompts
Four groups
Split the 25 prompts like this:
- 10 category prompts: “best [service] for [type of company]”
- 8 comparison prompts: one option against another for a specific job
- 5 problem prompts: the pain, described without your category name
- 2 brand prompts: your name, and what it does
Twenty-five is our recommendation, not a standard. It’s small enough to run by hand.
Wording rules
- Use the buyer’s words, not your marketing copy.
- Keep your brand name out of the 23 non-brand prompts. A mention only means something when the question is neutral.
- Name competitors only in comparison prompts. Below, they appear as [Competitor A].
- Add a US city, state or service area wherever a buyer would.
- Never edit a prompt in place. Retire it with a note and add a new ID.
Worked examples
Both examples are hypothetical and use placeholders.
Example (hypothetical): a regional HVAC contractor serving [metro]
- Category: “Who are the best HVAC companies for heat pump installation in [city, state]?”
- Comparison: “[Competitor A] vs [Competitor B] for furnace replacement in [city]”
- Problem: “One floor of my house is always much hotter than the other. Who should I call in [city]?”
- Brand: “[Your company] reviews”
Example (hypothetical): B2B scheduling software for field service teams
- Category: “Best scheduling software for a 20-person field service team in the US?”
- Comparison: “[Competitor A] vs [Competitor B] for dispatching technicians”
- Problem: “Our technicians keep double-booking jobs and we track everything in spreadsheets. What should we use?”
- Brand: “What does [Your company] do and who is it for?”
Step 2: Pick engines and run them the same way
Which tools
Choose from ChatGPT, Google AI Overviews and AI Mode, Gemini, Perplexity and Copilot. Pick three to five you can run monthly, and keep the list the same.
Bing Webmaster Tools has an “AI Performance” report covering Copilot and Bing AI citations. At the time of writing it was still labeled a preview (Previsible, 2026; eSEOspace, 2026).
The run protocol
- Run the audit in the same week each month, ideally by the same person.
- Open a fresh chat for every prompt, and sign out where you can.
- Run each prompt once per engine. No regenerating until you like the answer.
- Save the answer or a screenshot with the date on it.
Step 3: Score each answer
One row per prompt per engine.
| Date | Month | Engine | Prompt ID | Group | Prompt text | Mentioned (Y/N) | Linked (Y/N) | Competitors named | Sources cited | Source type | Facts about you correct? (Y/N/n.a.) | Notes |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
For “Source type,” use fixed labels: review or directory site, forum, video, LinkedIn, press, competitor site, your site. At month’s end, summarize mentions and links out of total answers, plus the most cited domains.
Step 4: Read month 1 against months 2 and 3
Treat month 1 as a baseline
Month 1 is where you stand, not a grade.
What counts as change
Compare totals by group and by engine, not single prompts. One prompt flipping from N to Y is noise (our view, not a statistic). Judge on three months together, and watch for new domains in “sources cited”.
Then put mentions next to form fills and booked revenue. A mention is not a lead; our piece on AI follow-up workflows for B2B covers that half.
Step 5: Fix what the “sources cited” column points to
Fix access first, because it blocks everything else. Then fix the source type that appears most in your sheet.
| If the cited sources are mostly… | Fix first |
|---|---|
| Your site is never cited | Check that Google Search Console and Bing Webmaster Tools show your key pages indexed. Make sure robots rules don’t block OAI-SearchBot, which OpenAI documents as the condition for appearing in ChatGPT search (eSEOspace, 2026). |
| Review and directory sites | Claim and complete the two or three profiles your buyers actually use. |
| Forums, video or LinkedIn | Take part honestly: answer questions, post useful things. |
| Competitor or publisher pages | Publish one plain page answering that exact prompt, with scope and prices. Keep structured data matching the visible text (Google Search Central, 2025). |
| Wrong or mixed-up facts about you | Write one plain sentence on what you do and for whom. Use it everywhere. |
Review sites: G2 found review-site citations were the strongest trust signal in chatbot answers, again for software buyers (G2, 2026). For design-agency queries, one review directory was the most cited domain in four of five models (Temso, 2026). That counts citations, not traffic.
Forums, video and LinkedIn: an index of more than 680 million citations lists Reddit, Wikipedia, YouTube, LinkedIn and Forbes among the most cited domains overall, and describes LinkedIn as dominant for B2B and executive queries (Everything-PR, 2026). The index owner has a PR agency partnership. Ahrefs data, as reported by Radiant Elephant (2026), shows brand mentions correlating with AI Overview visibility at about r=0.66, against about r=0.22 for backlinks. That’s correlation, not cause.
Consistent facts: G2 reported that consistent vendor descriptions across ChatGPT, Gemini and Claude read as a credibility signal (G2, 2026).
What not to fix first: llms.txt. Ahrefs data, as reported by PPC Land (2026), found 97% of llms.txt files got no requests in May 2026, and Google said it has no ranking effect.
Do it yourself or hand it off
Everything above is doable in-house. The hard part is repeating it monthly. If you’re weighing AI search optimization services, here’s how we handle it.
If you’d rather not run this every month, our SEO & GEO plan includes a monthly AI search visibility check. Every month we run the same fixed set of buyer prompts through the AI tools and log three things: whether your brand is mentioned, whether it is linked, and which sources get cited instead.
Checklist
- Write 25 prompts: 10 category, 8 comparison, 5 problem, 2 brand.
- Keep your brand name out of the 23 non-brand prompts.
- Pick three to five engines and keep the list fixed.
- Run each prompt once per engine, in a fresh chat, same week monthly.
- Log mentioned, linked, competitors, sources and source type.
- Judge across three months; month 1 is a baseline.
- Fix access first, then the most frequent source type.
If you are weighing what to pay for any of this, our marketing agency pricing 2026 comparison puts 17 published agency prices side by side, with ours in the same table.
FAQ
How is an AI visibility audit different from an SEO audit?
An SEO audit checks your site. An AI visibility audit checks what AI answers say about you and which sources they cite. It still builds on SEO basics: Google says being indexed and eligible for a snippet is enough for AI features, with no special schema or AI file required (Google Search Central, 2025).
Why 25 prompts, and why keep them fixed?
Twenty-five covers four types of buyer questions and is small enough to run by hand. Fixed prompts make change meaningful. The number is our recommendation, not a standard.
Can I see AI search traffic in Google Search Console?
Not separately. Traffic from Google’s AI features is included in the “Web” report (Google Search Central, 2025). Bing Webmaster Tools has an AI Performance report, labeled a preview at the time of writing (Previsible, 2026; eSEOspace, 2026).
Can anyone guarantee ChatGPT will recommend my business?
No. AI answers change, and nobody controls them. You control access, the facts you publish, the pages you offer and the mentions you earn, and you can measure the result each month.
Next step
To hand off the monthly check, see the SEO & GEO plan: $940 a month, 3-month minimum. It covers technical SEO upkeep, schema maintenance, one researched article a month and a monthly AI search visibility check, reported with rankings, organic traffic, leads and AI citations. Everything arrives in writing, with a recorded video when a screen helps. No meetings. Compare plans, including Lead Machine and Store Scale ($5,690 a month each). Want an outside look first? Ask for the free Growth Snapshot: a 7-minute video within 48 hours, including a quick check of your search and AI visibility. We don’t promise rankings, citations or traffic. Results not typical.
Sources
Everything-PR. (2026, October 20). AI platform citation source index 2026. https://everything-pr.com/ai-platform-citation-source-index-2026/
eSEOspace. (2026, October 13). Bing Webmaster Tools and IndexNow for AI search: What’s confirmed and what isn’t. https://eseospace.com/blog/bing-webmaster-tools-indexnow-ai-search/
G2. (2026, April 15). In the answer economy, don’t win the click, win the answer. https://company.g2.com/news/g2-research-the-answer-economy
Google Search Central. (2025, December 10). AI features and your website. https://developers.google.com/search/docs/appearance/ai-features
Moore, K. (2026, July 29). Your AI citations got you shortlisted with buyers. Now what? https://www.kaleighmoore.com/blog/2026/7/29/g2-evaluation-maze-report-findings
PPC Land. (2026, July 2). llms.txt adoption rises 8.8x but 97% of files get zero AI requests. https://ppc.land/llms-txt-adoption-rises-8-8x-but-97-of-files-get-zero-ai-requests/
Previsible. (2026, February 10). Bing AI performance report. https://ai.previsible.io/seo-ai-news/bing-ai-performance-report/
Radiant Elephant. (2026, July 14). Brand mentions now matter 3x more than backlinks for AI search. https://www.radiantelephant.com/brand-mentions-vs-backlinks-ai-search/
Temso. (2026, October 1). The sources AI trusts when recommending design agencies (Q3 2026). https://www.temso.ai/data/AI-Source-Citation-Patterns-in-Design-Agencies