Ahrefs AI Search Visibility Topics and ChatGPT Mentions

Ahrefs AI search visibility topics and ChatGPT mentions get mixed because both live in the same invoice. Brand Radar clusters a huge keyword-modeled corpus (260 to 405 million prompts). That is a topic research view. ChatGPT mentions on prompts sales hears are a ledger. We keep Ahrefs. We do not read Brand Radar as that ledger. The invoice can carry both, but the two products measure different things, and a deck that reads one as the other will mislead the client.

The mention log we add is Promptwatch. Topics and tags there are labels on prompts we typed (campaign, persona, country). Not an Ahrefs topic cluster. Essential is $95/mo. Explore is free, ChatGPT only, 10 prompts. Kick-off is $199/mo for more than one brand. The label is the part that lets a weekly table filter by campaign. A topic in Promptwatch is a label a human put on a prompt. A topic in Brand Radar is a cluster an algorithm built from a keyword corpus. Same English word. Different object.

Brand Radar is $199/mo per index plus an Ahrefs plan from $129/mo. Chatbot indexes refresh roughly monthly. An independent January 2026 test reported 3 ChatGPT mentions where 123 were verified. Custom prompts cost $50/mo per 2,500 checks. No Claude in native indexes. The January undercount is the number that tells you what Brand Radar is. It is a modeled index. A modeled index and a live ChatGPT session disagree, and the disagreement is the whole point. The modeled index is a research view. The live session is the ops number.

That last line is why we do not export Brand Radar topics and paste them into Promptwatch. Sales language is shorter and weirder than a keyword cluster. The January undercount is what you get when a modeled index and a live ChatGPT session disagree. We treat the live session as the ops number. A sales prompt is not a keyword. A keyword cluster is built from search demand. A sales prompt is built from what a buyer asks an assistant. The two corpora overlap, and they are not the same.

How we run both

Ahrefs stays the crawl and link system. Brand Radar stays if it is already paid. We pull it monthly and label it modeled. Weekly ChatGPT work uses a sales-language prompt list in Promptwatch. If ChatGPT cites a URL Ahrefs already knows, we use Ahrefs to target that domain. Promptwatch proves the model still uses it. The two products do different jobs in the same week. Ahrefs tells us about the link graph and the crawl. Promptwatch tells us about the answer the engine gave on Tuesday. The deck reads better when each product answers the question it owns.

How to get cited by ChatGPT is the edit list. Best AI SEO tools 2026 is the suite split.

Allow OAI-SearchBot. Do not pause rank tracking to make a GEO point. Canceling Ahrefs to "go GEO" is a bad trade. The trade is bad because Ahrefs still owns the crawl and the link graph, and GEO work still needs the link graph to target a domain the engine cited. A client who cancels Ahrefs to pay for a mention tool loses the crawl and keeps the mention. That is a net loss.

Semrush AI Toolkit at $99/mo per domain is the other modeled suite add-on. Same bucket as Brand Radar: useful if they already live there, not the mention ledger. The toolkit gives a modeled view of where the brand shows up in AI surfaces. It does not store the answer to a prompt you typed, and it does not store the cited URL.

Promptwatch paid checks run daily from the real UI. Essential includes citation and visitor analytics, but no listed crawler-log allowance. Agent Analytics starts on Professional for brand plans and is also included on the agency plans. We tag prompts by campaign so the weekly table can filter. A topic in Promptwatch is a label. A topic in Brand Radar is a cluster. Same English word. Different object. The label is the part that lets the weekly table filter by campaign. The cluster is the part that lets a researcher see demand across a corpus.

If they buy custom Brand Radar prompts, those checks are still modeled and metered. We trust a daily UI check on a named prompt for weekly tickets. Custom Brand Radar prompts are still Brand Radar. They refresh on the index cadence, not on a daily UI cadence. A weekly ticket that needs a daily answer should not read from a monthly index.

FAQ

Are Promptwatch topics the same as Brand Radar topics?

No. Same English word. Different object. A Promptwatch topic is a label a human put on a prompt. A Brand Radar topic is a cluster an algorithm built from a keyword corpus.

Do we still need Promptwatch if we buy custom Brand Radar prompts?

Those checks are still modeled and metered. We trust a daily UI check on a named prompt for weekly tickets. Custom prompts do not change the cadence of the index.

Do we cancel Ahrefs?

No. Links stay in Ahrefs. Mentions go to Promptwatch. The two products answer different questions on the same invoice.

What to do this week

Keep Ahrefs. Do not pause it. The crawl and the link graph are still in scope.
Write 15 ChatGPT prompts from sales, not from an Ahrefs topic export. Sales language is what buyers ask, not what a keyword cluster surfaces.
Tag them in Promptwatch by campaign. The tag is the part that lets the weekly table filter.
If Brand Radar is on the plan, pull it monthly and label it. Modeled data is useful when it is labeled as modeled.
Email hello@1001seomedia.com with the Ahrefs workspace note.