How to Measure Brand Visibility in AI

Brand visibility in AI is how often ChatGPT, Perplexity, Gemini, Copilot, and Google AI Overviews name your brand when buyers ask questions in your category.

You do not measure it with a Google rank. You measure it with a prompt set, a scan log, and a few rates that stay comparable week to week.

Referral traffic from AI is useful when someone clicks. Most of the effect happens before the click. The model names three tools. The buyer picks from that shortlist. If you were never named, analytics never sees the loss.

This guide covers the metrics that hold up, how to build the prompt set, and how to read those numbers in an AI SEO Tracker report.

Why rank tracking is the wrong scoreboard

Classic SEO reports a position. Everyone who searches that keyword sees the same list, more or less.

AI answers are not a list. The same prompt can name different brands on different runs, on different engines, and with search on versus off. Profound notes that ChatGPT's sources overlap only partly with Google's. You can rank well organically and still be missing from the answer.

Search Engine Land's guide on how to measure brand visibility makes the same split: clicks and impressions miss the mention that never became a session. A user can see your name in ChatGPT, then type your URL later. That visit looks like direct or branded search. The AI step is invisible unless you measured the prompt.

Google Search Console is still worth watching. In June 2026 Google added Search Generative AI performance reports for impressions in AI Overviews, AI Mode, and Discover. That is first-party data for Google's surfaces only. It does not tell you whether ChatGPT or Perplexity mention you.

You need a prompt-level scoreboard next to GSC, not instead of it.

The five metrics that actually travel

Keep the definitions tight. Vendors rename them. The underlying counts do not change.

1. Mention rate

Mention rate is the share of scanned answers where your brand name appears.

Mention rate = (answers that name you ÷ answers scanned) × 100

If you scan 40 prompts across 5 engines (200 answers) and you appear in 50, your mention rate is 25% for that set.

This is the baseline. Brainlabs calls the volume version of this "AI visibility volume." Profound calls a similar ratio a visibility score. Discovered Labs uses "citation rate" for a close formula (queries where the brand appears ÷ queries tested). Pick one name and do not mix the denominator.

Always state the set: which prompts, which engines, which dates. A 40% mention rate on 8 branded prompts is not the same as 40% on 40 discovery prompts.

2. Share of voice

Share of voice is your mentions relative to the competitive set in the same answers.

Share of voice = (your brand mentions ÷ all tracked-brand mentions) × 100

Mention rate says whether you show up. Share of voice says whether you own the category or just sneak in. You can appear in 20% of answers and still lose SOV if three competitors appear in 50%.

Use a fixed competitor list. If you add a new rival mid-quarter, the SOV trend breaks. For the classic formula outside AI, see how to calculate share of voice.

3. Citations vs mentions

A mention is your name in the answer text. A citation is a linked source, footnote, or Sources chip pointing at a URL (often yours, sometimes a third party talking about you).

HubSpot's AEO write-up draws the same line: mentions shape perception with no click. Citations are attributed sources.

Track them separately:

	Mention	Citation
Question	Did the model name us?	Did it use a URL?
Typical commercial prompt	"best X for Y" names 3-5 brands	Sources list 3-8 domains
Failure mode	Competitor shortlist, you absent	Model uses Reddit/G2, not your page
Fix	Entity + third-party co-occurrence	Extractable page + crawl access

You want both. A mention without a citation still puts you in the shortlist. A citation without a mention can send a click while the recommendation goes to someone else.

4. Scanned vs missing prompts

This is the operational metric most dashboards hide behind a single score.

• Scanned: prompts you ran this period, with a stored answer.
• Mentioned: scanned prompts where you appear.
• Missing: scanned prompts where you do not appear.

"Missing" is the work list. A blended 22% mention rate does not tell a writer which five questions to fix. The missing-prompt table does.

Also track unscanned prompts: items in the library that did not run (quota, engine error, or not yet scheduled). Do not treat unscanned as missing. Missing means you looked and were absent. Unscanned means you do not know.

5. Per-engine split

One blended number hides the real story. ChatGPT, Perplexity, Gemini, Copilot, and AI Overviews do not cite the same pages.

Report mention rate and SOV per engine, then a rollup. A brand that wins ChatGPT and loses Perplexity needs different sources, not a generic "AI SEO" rewrite. LLM SEO covers how source preferences differ by model. Measurement has to respect that split or you will optimize the wrong surface.

Optional extras, once the five above are stable:

• Position in the named list (first vs last among recommended tools)
• Sentiment of the sentence that names you
• Cited URL (which of your pages, or which third-party page)
• AI referrals (chatgpt.com, perplexity.ai, and similar) in analytics

Those refine quality. They do not replace mention rate.

Build a prompt set, not a keyword dump

The prompt set is the sample. A bad sample makes every metric lie.

Start from buyer questions, not from your homepage slogan. Profound's tracking guide starts the same way: employees, customers, and existing keywords, rewritten as conversational prompts.

Use six buckets (this is the mix most AEO audits use, including CiteMyBusiness-style checklists):

1. Discovery. Best [category]. Top tools for [job].
2. Use case. Best [category] for [industry / team size / stack].
3. Comparison. [You] vs [rival]. [Rival] alternatives.
4. Problem. How do I [job] without [pain]?
5. Proof. Is [you] worth it? [You] reviews.
6. Branded. What is [you]? [You] pricing.

Twenty to fifty prompts is enough to start. Past that, tag them so you can filter by intent. Do not mix branded and category prompts in one mention-rate headline. Branded should be high. Category is the competitive number.

Phrase them the way people talk to chatbots. "Best CRM" is a keyword. "Best CRM for a 5-person agency that already uses Gmail" is a prompt. The second is what you measure.

Re-run the same wording. If you rewrite the prompt every week, you cannot tell whether the model changed or you did.

Manual log, then automate

A spreadsheet still works for a first baseline.

Columns: date, engine, prompt, mentioned (Y/N), citation (Y/N), cited URL, competitors named, notes.

Run each prompt on ChatGPT, Perplexity, Gemini, Copilot, and Google (for AI Overviews). Save the answer or a screenshot. Repeat on a fixed weekday.

Limits of the manual log:

• Answers vary across runs. One check is a snapshot, not a rate.
• Logged-in accounts personalize. Use a clean session when you can.
• Five engines times 30 prompts is a half-day. You will skip weeks.

Wix's AI Search Lab roundup of tools that track brand visibility in AI search is a fair map of the dedicated trackers. They all do the same core job: send a prompt list to multiple models, parse mentions and citations, and chart the rates over time. Compare them on engines covered, whether you bring your own prompts, and whether you can export the raw answers. Our take on the category is in best AI SEO tools.

AI SEO Tracker is built around that prompt list. You add the questions, pick competitors, and scan ChatGPT, Perplexity, Gemini, Copilot, and AI Overviews. The report is the five metrics above, not a mystery 0-100 score with no denominator.

What you should see on a project report:

• Mention rate for the scanned set, with a per-engine breakdown
• Share of voice against the competitors you added
• Citations vs mentions so you can tell name-drops from sourced answers
• Scanned vs missing prompts as the actual backlog
• Sources the models used, so you know which sites to earn

A $49 audit is a one-time scan week (50 prompts). Ongoing tracking starts at $79/mo. MCP lets an agent read the same prompt, snippet, competitor, and scan data you see in the UI. That is the difference between a screenshot in Slack and a number you can trend.

Page Inspector, on Pro, shows how models read a URL. Use it after the report tells you which page should be winning a missing prompt.

How to read a week of numbers without fooling yourself

Separate branded from category. If branded mention rate is low, your entity is messy (conflicting descriptions, thin About page, missing profiles). If category mention rate is low, you are not in the consideration set. Those are different projects.

Do not celebrate a blended uptick. Check whether you scanned a different mix. Adding 10 easy branded prompts will lift mention rate without any real win.

Watch missing prompts, not only the average. Moving 4 of 12 missing discovery prompts is a real change. A 2-point SOV wiggle on a small sample is noise.

Read sources before you rewrite. If every citation for "best X" is a G2 listicle, a new blog post on your domain may not be the first lever. Get on that list. Generative engine optimization is the content half. Measurement tells you which half you are on.

Map rates to something commercial only after the rates are stable. Brainlabs is right that branded search, direct traffic, and AI referrals are validation, not proof by themselves. Overlay mention rate and AIO inclusion on those series. Look for lead/lag. You will not get prompt-to-purchase attribution from the vendors.

Cadence. Weekly scans catch drift. Daily scans on a tiny set mostly add noise. Revisit the prompt list quarterly, not every time a stakeholder has a new phrase.

Discovered Labs publishes working B2B SaaS targets of 10-15% citation rate on category queries as a starting point, with market leaders above 30%, and branded queries above 80%. Treat those as one team's yardstick, not a law. Your competitive set and prompt mix change the ceiling. Trajectory on a frozen set beats copying someone else's percentage.

A one-hour measurement setup

1. List 25 prompts in the six buckets above. Freeze the wording.
2. Add 3 to 5 competitors you actually lose deals to, plus the aggregators that steal citations (G2, Reddit, Wikipedia) if they show up.
3. Run a baseline across five engines. Save answers.
4. Compute mention rate, SOV, citation rate, and the missing-prompt list. Split by engine.
5. Pick five missing prompts. That is next week's content and outreach list.
6. Schedule the same 25 prompts weekly. Do not change the set for 30 days.

If you want that loop without the sheet, run the $49 audit or a trial on aiseotracker.com and export the missing prompts.

FAQ

Is mention rate the same as share of voice?

No. Mention rate is you vs the prompt set. Share of voice is you vs other brands inside the answers. You need both.

How many times should I rerun one prompt?

Enough to see a rate, not a coin flip. A single ChatGPT answer is a screenshot. Several runs across a week, on a fixed prompt, is a mention rate. Dedicated tools batch this. A sheet should at least rerun the core 10 prompts each cycle.

Can I measure this in Google Search Console alone?

No. GSC's generative AI reports cover Google's AI features (impressions, pages, country, device). They do not cover ChatGPT, Perplexity, Gemini chat, or Copilot. Use GSC for Google. Use a prompt scan for the assistants.

What if AI traffic is up but mention rate is flat?

You may be getting clicks from citations while the recommendation still names competitors. Or a few URLs are attracting crawler traffic. Read citations vs mentions, then the missing-prompt list. Traffic without mention rate is a leaky story.

Measure the prompts your buyers type. Report mention rate, share of voice, citations vs mentions, scanned vs missing, and the per-engine split. Everything else is a dashboard skin on those counts.