Home→Services→AI visibility monitoring

AI visibility monitoring and monthly GEO reports

Every month I run a fixed set of tracked prompts through the AI systems your buyers use, with the same method and on the same dates. I track mentions, citations and AI referral traffic, and check that past fixes on your site have not been rolled back. You get a report I write myself, with the change since last month and a plan for the next one.

Fixed method and check dates 50–200 tracked prompts Up to 7 AI systems Report and next-month plan
Diagram: prompt coverage month by month, with a drop marked in month 5020%40%60%month 1month 3month 5month 7month 5: a model update, coverage droppedsame prompt set as before, so the drop is realprompt coverage: 12% → 44% in eight months
One check on its own says little. The drop in month 5 can only be read against the four comparable checks that came before it.

One dot shows where you stand. A trend takes at least three.

When do you need monitoring?

When some work is already done and you want to know whether it is working.

The work is done, the effect is unknown

The audit is finished and the fixes are live, but nobody checks whether the answers changed.

Fixes roll back unnoticed

After a site release, the markup is gone and key pages are blocked to crawlers again.

Competitors keep moving

New sources and competitor pages start getting cited in your market, and you are the last to find out.

One answer proves nothing

Ask once and you are named, ask again and you are not. Without a fixed method, you cannot tell a real change from chance.

AI traffic is not separated

Visits from AI systems blend into general referral traffic, and the channel looks as if it does not exist.

Nothing to show management

Things feel better, but you have no number to put in front of management.

What gets tracked
every month?

Prompt set
  • An agreed set of tracked prompts, kept up month after month
  • New prompts added when you launch services or enter new markets
  • Existing prompts never reworded, so the months stay comparable
  • Separate groups for commercial, brand and local questions
Regular checks
  • The full set run in the agreed AI systems on the same dates each month
  • Each prompt run several times; the report counts the share of runs that mention you
  • Brand mentions, domain citations and your position in lists
  • Raw answers saved for later comparison
Competitors and sources
  • Who gets named instead of you, and how often
  • New sources that start appearing in citations
  • Competitor pages that start showing up in answers
  • Share of voice across competitors – Standard and Extended
Fact accuracy
  • Whether AI systems state your prices, locations, services and contacts correctly
  • New inaccuracies flagged as they appear
  • Answers checked against a reference list of verified facts about your business
  • A flag when an error needs separate work
Analytics and traffic
  • A separate segment for AI referral traffic in GA4 or your analytics
  • Visits from ChatGPT, Perplexity, Gemini, Copilot and other AI sources over time
  • Google Search Console and Bing Webmaster Tools data
  • AI visits linked to inquiries and leads where your analytics allows it
Regression checks
  • Markup, redirects and page access still in place after site releases
  • AI crawler access rechecked after server or CDN changes
  • Priority pages reviewed after each release you tell me about
  • An early warning when something breaks
Report and plan
  • A monthly report with the change in every metric
  • New problems and opportunities
  • Hypotheses for the next period, with reasons
  • A task list for next month

How does the project run?

Method setup

You approve the prompt set, the AI systems and the check dates, and they stay fixed. 2–3 days.

Baseline

The first run becomes the reference point for every later comparison. 2 days.

Analytics setup

AI referral traffic gets its own segment in your analytics. 1–2 days.

Monthly check

The same method on the same dates each month; every two weeks on Extended.

Report

What changed, new problems and opportunities, and what I think caused them.

Next-month plan

What to do next month and why those tasks come first.

What do you get each month?

The tracking sheet keeps every check since the start, so after six months it shows the direction of change for each prompt.

XLSX

Tracking sheet

Every check since the start: prompts, systems, mentions, citations and dates.

PDF

Monthly report

What changed this month and since the start, with my reading of why.

XLSX

New issues and opportunities

Inaccurate facts, newly cited sources and competitor activity found during the month.

DOCX

Next-month plan

Priority tasks with reasons, and who is responsible for each.

LINK

Dashboard

Extended only: live access to current data and alerts when visibility drops.

MSG

Questions in writing

Questions about the report answered by email or messenger within one business day.

I keep the report short. Ten pages would mean I measured the wrong things.

Schedule and format

  • Checks on the same dates each month, so the data stays comparable
  • Report within 3 business days of the monthly check
  • Billed at the start of each month
  • Plan on at least 3 months: one check shows where you stand, and a trend takes several

What I need from you

  • Read access to analytics, Google Search Console and Bing Webmaster Tools
  • A heads-up about releases and changes on the site
  • Current prices and services to check AI answers against
  • Feedback on the hypotheses for next month

Not included

  • Fixes on the site: pages, markup, content
  • External placements and outreach
  • Growth in visibility – monitoring measures it, and the work that changes it is in the monthly plans

Which numbers are tracked?

The same metrics every month: being able to compare one month with the next matters more than measuring everything.

Prompt coverage

The share of prompts in the set where your brand appears in the answer.

Domain citations

How often your site appears among the sources an answer cites.

Position in lists

Where you appear when an answer names several companies.

Share of voice

How often you are named compared with competitors – Standard and Extended.

Fact accuracy

The share of answers that state your prices, services and contacts correctly.

AI referral traffic

Visits from AI systems and the inquiries that follow them.

A Telegram store builder named by 4 of 4 tracked AI systems

Before the work, none of the 4 tracked AI systems named the service when asked how to start an online store in Telegram. It took 90 days to appear in answers; now all 4 name it, and ChatGPT lists it first of 8. Russian-speaking market.

Read the case study

Questions about monitoring

AI answers change every time. What exactly do you measure?
Frequency. Each prompt runs several times, and the report counts the share of runs that mention you. The method, the prompt set and the check dates stay fixed – without that, a change from one month to the next means nothing.
Will monitoring improve my visibility?
By itself, no. Monitoring is measurement, analysis and reporting. The monthly GEO plans include tracking too and add the work: technical fixes, page optimization, content briefs and external placements. If you need both the numbers and the changes, start with a plan.
Why pay for this if tracking software costs $29 a month?
Tracking tools cost from $29 to $500 a month. They give volume and regularity, and I use them too. What they leave to you is reading the answers: how your brand is described, who is named before you, which sources get cited and what to change on the site. That reading and the plan that follows are what the monthly fee covers.
Which AI systems do you track?
Basic covers 4, Standard 6 and Extended 7. The full list is ChatGPT, Google AI Overviews and AI Mode, Gemini, Perplexity, Microsoft Copilot, Claude and DeepSeek. The systems are chosen at setup, based on where your buyers look for answers.
Can I start monitoring without an audit?
Yes. The first check then becomes the baseline. The prompt set has to be built from scratch, so the first month goes into setting up the method, and change analysis starts from the second.
What if nothing changes in a month?
That is a result too, and the report shows it as is. A flat month means a hypothesis did not work or the changes have not reached the models yet. I do not dress it up, because next month’s plan is built on these numbers.

Discuss monitoring

Tell me what has already been done on your site, and I will suggest a prompt set and a reporting format.

I reply within one business day. For a larger project, fill in the project brief.