AI Brand Monitoring: What to Track Every Month, and the Changes Worth an Alert
AI brand monitoring means asking AI platforms the same questions every month and recording what changed: whether they name your brand, where they put it, whether they link to your site, and whether what they say is true. One audit tells you where you stand. The second tells you what moved.
Below: why answers change, what to track, which changes need an alert, and how to choose an AI brand monitoring tool. The example is our AI visibility audit of Nike, Adidas, Puma and Under Armour, run on 15 October 2026.
Why AI answers about your brand change
Three kinds of change, each with a different fix.
1. The vendor ships a new model
OpenAI's model docs: "Snapshots let you lock in a specific version of the model so that performance and behavior remain consistent." When the version changes, behaviour can too, so store the model name and date with every answer.
2. An answer from memory or from search
In our audit, ChatGPT (gpt-5.6-luna) and Claude (claude-sonnet-5) answered from memory, with web access off. Perplexity (sonar) searched the live web.
A memory answer is bounded by the model's knowledge cutoff: OpenAI lists gpt-5.6-luna's as 16 February 2026. Nothing you publish this quarter reaches an answer a model gives from memory. That changes with a new model, not with your site.
A search answer can change any week, because it reads live pages. You cannot see the searches it runs either: OpenAI says ChatGPT search "typically rewrites your query into one or more targeted queries". Your prompts are a sample of what buyers see, not a copy.
3. Crawler access changes at your CDN or firewall
A search-based platform can only cite pages its crawler reaches. OpenAI asks you to "allow OAI-Searchbot to crawl the site and confirm that the website host or content delivery network allows traffic from OpenAI's published searchbot IP addresses." Sites opted out of OAI-SearchBot "will not be shown in ChatGPT search answers".
The second gate can close without anyone touching robots.txt. Cloudflare announced that from 15 September 2026, "for all new domains onboarding to Cloudflare, the categories of Training and Agent will be blocked by default on the pages that display ads". See Cloudflare's AI crawler default change.
What to track on each platform
One row per prompt, per platform, per month. Do not blend platforms into one number: memory and search answers move for different reasons, and an average hides which one moved.
| Track | Record | Why |
|---|---|---|
| Mentioned | Were you named at all? | Not named means not considered |
| Named first | The first brand named | Mentioned and recommended are different results |
| Named before you | Every brand ahead of you, in order | Shows who you lose to, prompt by prompt |
| Your site linked | Did a search-based answer link your domain? | Memory answers link nothing |
| Facts right | Each claim against your own fact list | Our audit's errors sat in facts that had changed |
| Model, date, memory or search | Logged with every answer | Without them, months do not compare |
For how these roll up into a score, see the GEO metrics that matter. To act on "named before you", see competitor GEO analysis.
The four changes worth an alert
Most movement is wording: Claude gave PUMA's CEO a different start year in two answers on the same day. These four changes need action.
1. A new wrong fact
Trigger: an answer contradicts a fact on your list. First check: memory or search? Nothing you publish fixes a memory error before the next model, so log it and remove the old version from your own pages. A search error came from a page: find the page.
2. A competitor overtakes you in "named first"
Trigger: on a prompt you led last month, a competitor now leads on the same platform. First check: did the model change? If not, and the platform searched, compare its cited sources with last month's.
3. A search-based platform stops citing your site
Trigger: a platform that searched linked your domain last month and does not now. First check: access before content. Confirm the page still exists and its crawler can still reach it.
4. An AI crawler is newly blocked
Trigger: a retrieval crawler that reached your page last month now gets a 403 or a challenge page, or robots.txt now disallows it. First check: what changed in your CDN, firewall or bot settings. A test request that borrows a crawler's name does not come from its published IP addresses, so treat a block as "a rule matched" and go read the rule.
What our sportswear audit would give a monitor
One run is a baseline. This is month one on ChatGPT, Perplexity and Claude; Gemini is left out because a token limit on our side cut its answers short.
- Wrong facts. Claude said Under Armour "owns popular fitness-tracking apps" (MyFitnessPal, MapMyRun). Under Armour sold MyFitnessPal in December 2020, and Outside Inc. acquired MapMyFitness, including MapMyRun, in August 2024. Claude also dated PUMA's CEO change to 2023 in one answer and 2024 in another; PUMA lists Arthur Hoeld as CEO, on its board since July 2025. All 16 CEO answers named the right person. We found these errors by hand.
- Named first. Nike led 13 of 15 unbranded answers. Under Armour led none, and Nike came before it in 8 of the 9 answers that named it.
- Site linked. Perplexity gave 75 citations across the five unbranded questions. One linked to nike.com. None linked to the other three brands.
- Crawler access. All 12 retrieval crawler user agents we tested reached puma.com. On adidas.com and underarmour.com even our ordinary browser request was refused (HTTP 403 and HTTP 418), so those two checks measured nothing. Record that as "not measured", never "blocked".
In a different answer, the same Claude model wrote that "MyFitnessPal was sold in 2020". The error changed with the question, so a monitor re-asks a fixed set of questions every month.
A monthly AI brand monitoring routine
Our weekly GEO review routine is a 15-minute check for anything urgent. The monthly pass compares every row with last month's.
- Update your fact list first: prices, leadership, products, ownership. You cannot check an answer against a fact you have not written down.
- Re-ask the same prompts on the same platforms. If you change the prompt set, start a new baseline.
- Log the model, the date, and memory or search for every answer, and check the vendors' model pages: a new model is when memory answers can move.
- Diff the rows and mark each change as one of the four alerts, or as noise.
- Re-run the crawler access check and compare what robots.txt allows with what each crawler gets back.
How to choose an AI brand monitoring tool
Most results for "AI brand monitoring tools" are ranked lists. Check any tool against what monitoring needs:
- Platforms covered, and memory or search per answer. Without that, you cannot tell which fix applies.
- Stored raw answers. You need the sentence that was wrong, not only a score that dropped.
- Dated model versions on every answer.
- Competitors measured on the same answers. Separate samples make "named before you" meaningless.
- Accuracy checks against your facts. Ask what it does with a claim your facts do not cover.
- Crawler access checks. Reading robots.txt is half the job; it should also request your pages as each AI crawler.
- Gaps shown as gaps. A failed answer is "not measured", not "not mentioned".
- Price against prompts × platforms × months, with the competitors you need.
No markup, file or rewrite has credible evidence that it makes AI favour you. The levers that hold up are being reachable and being right. See how LLMs prioritise information about brands.
99Visibility audits ChatGPT, Gemini, Perplexity, Google AI Overviews and Claude, tracks your competitors, checks answers against the brand facts you enter, and checks whether AI crawlers can reach your site. Your first audit is free with no credit card and covers ChatGPT and Google AI Overviews. See the paid plans. Inside the dashboard, Agent V reads your audits and your site and drafts the fix, such as page copy for you to paste. It never edits your site itself.
FAQ
What is AI brand monitoring?
Asking AI platforms the same questions on a schedule and recording whether they name you, in what order, with which links and which facts. Each run is checked against the last.
How is it different from regular brand monitoring?
Brand monitoring tracks what people publish about you. AI brand monitoring tracks what AI says when a buyer asks, which can be wrong even when everything published about you is right.
How often should you check what AI says about your brand?
Monthly for the full comparison, with a quick weekly look for anything urgent. Re-check after any model release.
Start your baseline today. Run a free AI visibility audit, then run it again next month and compare.
Want more of our research in Google Search and AI Mode? Add us as a preferred source.
Get your AI visibility score
One number from 0 to 100: how often AI mentions you, how accurately, in what tone, and whether it cites you, scaled by the market weight of the platforms that name you.
Start Free Audit46/100
Named on 3 of 5 platforms× 0.76 coverage
Related Articles
We Ran an AI Visibility Audit on Nike, Adidas, Puma and Under Armour. Here's What AI Got Wrong
We asked ChatGPT, Perplexity and Claude the questions sportswear shoppers ask, then checked the answers by hand. Nike was named first in 13 of 15 answers. Under Armour was named first in none, and Claude says it still owns MyFitnessPal.
AI Visibility Monitoring: A 15-Minute Weekly GEO Review
A fixed 15-minute weekly routine for AI visibility monitoring: crawler access, answers, first-party data, and a triage table for when something changes.
Should I Block AI Crawlers on Cloudflare? A Per-Bot Decision Guide
Block training crawlers if you object to training; keep search and agent crawlers open to stay in AI answers. What Cloudflare changed on September 15, 2026, per crawler, and how to check yours.