AI Visibility Monitoring: A 15-Minute Weekly GEO Review

AI visibility monitoring means re-running the same fixed set of prompts on the same AI platforms on a schedule, and watching three things: whether you are named, whether what is said about you is correct, and whether AI crawlers can still reach your site. Fifteen minutes a week is enough if the prompts, platforms and settings stay fixed. Change any of them and your trend line means nothing.

Below is the setup you do once, the 15-minute weekly review, a triage table for when something changes, and how to tell a real change from noise.

What to monitor, and why each one changes

Most of what moves your AI visibility happens without you touching anything: a model release, a crawl, a firewall rule, a competitor's new page. That is why monitoring beats a one-off audit.

The last row is deliberate. Google states that third-party tools don't have access to our internal ranking data, and no AI platform publishes a position you could check a tool against. Monitor whether you are named, cited and correct, not where.

Set it up once (about an hour)

1. Write a fact sheet. Prices and plans, founding year, headquarters, CEO, ownership, main features, and what you no longer offer. This is what "correct" means in every review.
2. Pick your prompts. Ten for the weekly set (two pricing, three category, two comparison, two brand, one problem-solving) and 20 to 40 for the monthly set. Write them down word for word.
3. Fix the conditions. Logged out, same location, same mode for each platform. OpenAI notes that with memory on, ChatGPT may use saved memories when rewriting a search query, so your own account is a moving target.
4. Take a baseline. Run every prompt three times. The spread you see is your noise level.
5. Save the access baseline. A copy of robots.txt and the status code each AI bot gets on your pricing page.
6. Connect the first-party sources. Search Console, Bing Webmaster Tools and GA4, each verified for the domain you are monitoring.
7. Set up the log. The free AI visibility tracking template has sections for scores, an accuracy log and competitors.

The 15-minute weekly review

Minutes 0 to 3: access

Access is first because one broken gate explains every other drop. Open your robots.txt and compare it with last week's copy. Then request your pricing page as the main search bots from a terminal (on Windows, open PowerShell and type curl.exe instead of curl, because plain curl is a different command there):

curl -sIL -A "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36; compatible; OAI-SearchBot/1.4; +https://openai.com/searchbot" https://www.yoursite.com/pricing

curl -sIL -A "Mozilla/5.0 (compatible; PerplexityBot/1.0; +https://perplexity.ai/perplexitybot)" https://www.yoursite.com/pricing

The user-agent strings are the ones OpenAI and Perplexity publish. If you see a 301 or 308 first, the last status code is the one that counts. A change from 200 to 403, 429 or a challenge page is your most urgent finding of the week. Because the request comes from your IP rather than the bot's, confirm it in your CDN logs before you rewrite rules.

If you are on Cloudflare, open the AI bot settings too (Security Settings, then Configure AI bot policies). Our guide on whether to block AI crawlers on Cloudflare explains what each option now blocks.

Minutes 3 to 9: answers

Run the ten weekly prompts, or open the latest results in your tool. For each answer, mark three things against last week: named or not, facts right or wrong, and whether it links to you. Pricing prompts first, because a wrong price costs you a sale this week.

Minutes 9 to 12: first-party data

• Search Console: the Generative AI performance report shows impressions in AI Overviews and AI Mode by page. It has no clicks or position, so watch which pages gained or lost impressions.
• Bing Webmaster Tools: AI Performance shows citations across Copilot and Bing's AI summaries, plus grounding queries, the phrases the AI used to retrieve your content. New grounding queries are the best source of new prompts you will find.
• GA4: sessions in the AI Assistant channel. It excludes AI Overviews and AI Mode, which GA4 files under Organic Search, so it is always an undercount.

Minutes 12 to 15: log and triage

Write the week's numbers into the log, then run every change through the table below. Anything marked "today" gets an owner before you close the sheet.

What to do when something changes

For the full correction process on wrong answers, see how to fix ChatGPT's wrong information about your company.

How to tell a real change from noise

• Repeat before you react. The same prompt can return a different answer on the next run. Count a change as real when it shows in at least two of three runs, two weeks running.
• Compare like with like. An answer from memory and an answer after a web search are different events. OpenAI says ChatGPT may search the web automatically when a question would benefit from current information, so log which mode each answer used.
• Log model releases. A new model can change memory answers overnight. OpenAI's model pages list each model's knowledge cutoff, for example April 20, 2026 for chat-latest as of October 2026.
• Log your own changes. Price changes, redesigns and redirects can cause drops that look like an algorithm change.
• Do not count a missing Overview as a loss. Google says AI Overviews often don't trigger. No Overview is Google's choice, not a ranking drop.

Monthly and quarterly add-ons

Monthly (45 minutes): run the full prompt set, score sentiment, and compare your AI share of voice with the competitors you track. Our guide to GEO metrics defines each number and how to report it.

Quarterly (one hour): re-check the fact sheet against reality, review your training-bot decisions, and refresh the prompt set from Bing's grounding queries and your sales calls. A new prompt set starts a new baseline, so change it on a fixed date and say so in the log.

Manual routine or a monitoring tool?

Before you buy, ask any vendor four questions:

1. Does it say which mode (memory or web search) each answer came from?
2. Does it test crawler access, including firewall blocks?
3. Does it check answers against your facts?
4. If it reports a rank, what is that rank validated against?

Our comparison of the best GEO tools applies those questions to the main vendors, and our AI visibility tracking page lists what 99Visibility monitors. Its Starter plan covers 15 audits a month across all six platforms.

Frequently asked questions

How do I monitor AI visibility?

Fix a set of prompts, platforms and conditions, then re-run them on a schedule and record whether you are named, whether the facts are right and whether you are linked. Add a weekly check that AI crawlers can still reach your site, and read Search Console's AI report, Bing's AI Performance report and GA4's AI Assistant channel alongside it.

How often should I check my AI visibility?

Weekly for crawler access and your most important prompts, especially pricing. Monthly for the full prompt set and competitors. Re-check immediately after you change prices, change CDN or firewall settings, or see a new model announced.

Is there an AI visibility rank tracker?

Several tools sell one, but there is nothing to validate it against. Google says third-party tools have no access to its internal ranking data, and the order of brands in a generated answer shifts between runs. Track mention, citation and accuracy rates instead.

How do I track my company's visibility across AI search platforms?

Run the same prompts on each platform and log the results separately, with the mode each answer used. ChatGPT and Claude can answer from memory, Perplexity searches the web, and AI Overviews appear only when Google decides to show one, so a single combined number hides more than it shows.

What should a weekly GEO report include?

The status of each AI crawler, the week's answers on your fixed prompts (named, correct, linked), changes in Search Console AI impressions and Bing AI citations, and AI Assistant sessions in GA4. List every change with an owner and a due date, and note any model release or site change that week.

To run the weekly answer and access checks across all six AI platforms without doing them by hand, start a 99Visibility audit (all six platforms on paid plans; the free plan covers two). It checks answers against your facts and crawler access.