AI search visibility tracking your team can read

Tracking repeats an agreed set of buyer questions on agreed platforms, on a schedule, and records each answer with its date, sources and whether your business was named. Every result is reported with its denominator, such as “named in 4 of 12 completed checks,” so your team can see what changed and how much to trust it.

What one observation looks like

Question“Which Calgary mortgage brokers work with self-employed buyers?”Platform: AI assistant, web search on · Date: 18 Nov 2026 · City in question: Calgary · Run 2 of 3

“…brokers that list self-employed programs include Example Brokerage, which says it reviews two years of business income… [1]”

[1] directory.example/brokers/example-brokerage

  1. Named: yes, in the text.
  2. Cited URL: a third-party directory, not the brokerage’s own page.
  3. Accuracy: directory lists old hours. Correction requested; retest next month.
One dated observation, annotated (illustrative data)The record keeps the exact question, the platform and mode, the date and city, an excerpt of the answer, whether the business was named, and each cited URL. Here the answer names the business but cites a directory page with outdated hours rather than the business’s own location page, so the finding is “named, cited third-party source, hours inaccurate”.

What we agree before the first run

Tracking set-up
SettingExample
Questions15 questions, split into unbranded discovery and branded accuracy checks
Platforms and modesChatGPT Search, Perplexity, Google Search (AI Overviews when shown)
Market and languageCalgary, English; Montréal, French
ScheduleMonthly, in the first week
ReportResults by question and platform, the change log, and anything that needs review

Why answers vary, and how we handle it

The same question can produce a different answer an hour later, in another account or another city. We reduce noise by fixing the wording, recording the mode and location, running repeat checks where the scope allows, and keeping inconclusive results instead of dropping them. A trend is only reported when the denominator supports it.

Platform reports we can include

Where you authorize access, we add Google Search Console data (including the Generative AI performance report for AI Overviews and AI Mode, when your property has it) and Bing Webmaster Tools’ AI Performance report. Each is labelled with the product it covers. Neither shows ChatGPT results.

Questions buyers ask

Which platforms do you review, and how are results measured?

We agree the platforms with you. Coverage can include Google Search and its AI features, ChatGPT, the Gemini app, Perplexity, Claude and Microsoft Copilot. Every test records the exact question, platform, date, how it was run (consumer app, API or manual review), whether your business was mentioned, and whether your site was cited. Mentions and citations are recorded separately. Where you give us access, we also report Search Console data, referral visits from AI platforms and enquiries. We don’t use a made-up composite score.

How to measure AI visibility
How long does improvement take, and what affects the timeline?

New and fixed pages can be crawled within days or weeks. Visible changes in rankings and AI answers usually take a few months. The main factors are how competitive your services and city are, how much needs fixing, how quickly changes get published on your site, and whether other sites already describe you accurately. We set retest dates at the start so you can see progress rather than guess.

How we measure progress

Primary sources

Checked on 18 November 2026. Platform behaviour changes; we re-check these references when we revise the page.