Your firm appears in an AI answer. Good. But did a buyer see it, trust it or contact you? Those are three different questions.
AI search visibility measures how often your business is mentioned or cited for a defined set of buyer questions. To make it useful, record the questions, engines, markets and repeat runs behind the number, then connect that evidence to enquiries and customers. A visibility score alone cannot tell you whether the work is paying back.
The useful bits, at a glance
- Keep AI mentions, citations, enquiries and revenue as separate measures.
- Record the prompts, platforms and repeated checks behind a visibility score.
- Combine observed behaviour with what buyers say influenced them.
In this guide
- Why is AI search visibility hard to measure?
- What should you actually measure?
- How do you build a prompt panel?
- What free data do the platforms give you?
- How do you connect visibility to revenue?
- What a realistic first month can deliver
- A worked example: build a score you can actually explain
- Checklist
- Frequently asked questions
- Get a baseline
Why is AI search visibility hard to measure?
Three reasons.
Answers can change between runs, even when the question appears unchanged. Save the prompt, platform, settings, date and full answer, then repeat the check. A single result is an observation rather than a stable ranking.
Google does not separate it out. Google's documentation on AI features says traffic from AI Overviews and AI Mode is included in the "Web" search type in Search Console's Performance report. You cannot filter it on its own.
People see you without clicking. Pew Research found that when a Google AI summary appeared, users clicked a link inside it in just 1% of visits. If your name is in the summary, that exposure mostly will not appear in your analytics.
Use several measures and state the limits of each. Recorded AI answers, referral visits and customer-reported discovery answer different questions.
What should you actually measure?
| Measure | What it tells you | Where it comes from | How often |
|---|---|---|---|
| Share of answers | How often you are named for your buyer questions | Your own prompt panel, or a tracking tool | Monthly |
| Cited pages | Which of your pages AI systems rely on | Bing AI Performance report, manual checks | Monthly |
| Grounding queries | The searches AI used when it cited you | Bing AI Performance report | Monthly |
| AI referral visits | People who clicked through from ChatGPT and others | Google Analytics (utm_source=chatgpt.com) | Weekly |
| Self-reported source | New clients who say they found you via AI | Enquiry form and intake call | Every enquiry |
| Revenue from those clients | Whether any of it matters | Your finance team | Quarterly |
The first four tell you whether you are visible. The last two tell you whether it is worth anything.

How do you build a prompt panel?
This is the core of AI search visibility tracking, and you can do it with a spreadsheet.
- Write 20 to 40 real buyer questions. Take them from enquiry emails, sales calls and Search Console queries. Mix broad ("best employment solicitors in Bristol") with specific ("can my employer change my contract without consent?").
- Fix the list. If you change the questions every month you cannot compare.
- Run each question in each tool. ChatGPT with search on, Google AI Mode, Google search (to see if an AI Overview appears), Perplexity and Copilot.
- Run each more than once. Because answers vary, three runs per question per tool gives a truer picture than one.
- Record three things. Were you named? Was your site cited? Who else was named?
- Calculate share of answers. Times you were named, divided by total answers. Track it by tool and by service line.
A tracking tool can help repeat and record checks. Start with a manageable sample that the team can inspect directly. Reading the saved answers helps reveal inaccurate descriptions and irrelevant citations that a single score can hide.
What free data do the platforms give you?
Bing Webmaster Tools
Microsoft launched an AI Performance report in public preview in February 2026. It covers Microsoft Copilot, AI summaries in Bing and some partner integrations, and shows total citations, the average number of your pages cited per day, the grounding queries the AI used, and citation counts for individual URLs. It is the closest thing to first-party AI visibility data I know of, and many firms have never opened it.
Google Search Console
No AI-specific filter, but still useful. Watch impressions for the pages you are trying to get cited, and look for long, question-style queries. Google has said early AI Mode users asked questions two or three times longer than traditional searches, so growth in long queries is a reasonable sign.
Google Analytics
OpenAI's publisher FAQ says ChatGPT automatically adds utm_source=chatgpt.com to referral links. Set up a channel group or saved report that groups AI referrers together: chatgpt.com, perplexity.ai, copilot and gemini referrals. Volumes will be small. That is fine.
How do you connect visibility to revenue?
Connect visibility observations to the business’s enquiry and customer records.
Add a required "How did you hear about us?" field to every enquiry form, with "ChatGPT or another AI assistant" as an option, plus a free-text box. Ask the same question on the first call. Then, once a quarter, match won clients back to their source with your finance team.
Judge the work using evidence the business can check: qualified enquiries, customer records and invoiced revenue. Keep changes in rankings or clicks separate from those commercial outcomes. Our results page explains the evidence behind published Pieo engagements.
What a realistic first month can deliver
A first month can establish the baseline, agree the priority pages and scope, implement approved changes, and review what has been delivered. The next 90-day plan should reflect the evidence. Rankings, AI citations and leads cannot be promised on a 30-day deadline. See how the Pieo pilot works.
A worked example: build a score you can actually explain
Imagine an employment-law firm serving UK technology companies. This is an illustrative measurement plan, not a Pieo client result. Start with 20 questions drawn from sales calls: choosing an adviser, comparing retainers, handling a specific employer problem and assessing specialist experience.
Run each question three times in the same search-enabled engine, in the same market, over a defined window. That creates 60 observations. Save the full answers, dates, source links and engine settings. Do not quietly rewrite the prompts after seeing disappointing results.
If the firm is named in 12 answers, its mention rate for that sample is 20%. If its own website is linked in six answers, its citation rate is 10%. Those numbers describe this panel. They do not mean the firm reaches 20% of all AI users.
| Observation | Example result | Decision it supports |
|---|---|---|
| Named in answers | 12 of 60 runs | Check which buying questions include the firm. |
| Own pages cited | 6 of 60 runs | Identify which pages support the answer. |
| Wrong service description | 3 of 12 mentions | Correct inconsistent service and profile information. |
| Buyer-reported discovery | Recorded separately in the CRM | Learn whether visibility contributes to real conversations. |
Keep an explicit “no search result” or “no answer” category. Report how you treat it in the denominator. Keep branded questions separate: asking “is this firm good?” is a different test from “which firm should I hire?”
Repeat the panel next month and compare the underlying questions, not just the headline percentage. A jump driven by one volatile query is weaker evidence than repeated gains across several commercially relevant questions. For a small sample, directional change is more useful than false precision.
The useful question is: which buyer questions are we becoming a credible answer to, and what commercial evidence accompanies that change?
- Visibility
Mentioned or cited
- Enquiry
A real buyer gets in touch
- Revenue
A customer is won
A practical framework from Pieo
Checklist
- Fixed list of 20 to 40 buyer questions
- Monthly runs across ChatGPT, AI Mode, AI Overviews, Perplexity and Copilot
- At least three runs per question per tool
- Bing Webmaster Tools verified, AI Performance report checked
- AI referrals grouped in Google Analytics
- "How did you hear about us?" on every enquiry route
- Quarterly revenue match with finance
Frequently asked questions
Is there a Google Search Console report for AI Overviews?
Not at the time of writing. Google counts AI feature traffic inside the normal "Web" search type.
Which AI search visibility tool should I use?
Any reputable tool that lets you fix your own prompts, runs them repeatedly and exports raw answers. Avoid tools that give a single score without showing you the answers behind it.
How many prompts do I need?
For a firm with a few service lines, 20 to 40 is enough to spot trends. More prompts add precision but also noise if they are not questions real buyers ask.
My AI referral traffic is tiny. Does that mean it isn't working?
Not on its own. Most people who see your name in an AI answer do not click. Look at share of answers, branded searches and what new clients tell you.
Get a baseline
Want more revenue from Google and ChatGPT? Enter your website at our free Search Check. Our team reviews every site, including how you show up in AI answers, and you hear back within one working day.


.jpeg)
%2520(1).jpeg)
