Semrush AI Toolkit Topic Research for AI Visibility and ChatGPT

Semrush AI Toolkit topic research is the request that shows up after the retainer already includes Semrush. Someone wants AI visibility and ChatGPT coverage without a second vendor. The Toolkit is $99/mo per domain on modeled prompts. We keep Semrush. We do not treat those modeled topics as the ChatGPT editorial list, because they are not the same object even though the word "topic" sits in both products. This post is the topic-research cut. The Toolkit-as-monitor cut is a different post.

We still read Google's AI features page for Overviews. It does not store ChatGPT or Perplexity answers. That is the recurring lesson of this whole category. Google's docs cover Google's surfaces. The assistants have their own log, and the log lives somewhere else.

The prompt list we actually write from is Promptwatch. Topics and tags there are labels on questions sales already hears. Query fan-outs add the related searches the model ran, which is a different kind of expansion from what a keyword tool produces. Essential is $95/mo. Explore is free, ChatGPT only, 10 prompts. Kick-off is $199/mo with unlimited projects and 10 seats.

What the Toolkit is good for

Semrush stays the keyword and rank system. Position Tracking, content-gap URLs, who already ranks in Google. The Toolkit sits on that graph. If the team already lives in Semrush, adding the AI column is cheaper politically than opening procurement, and we respect that. Procurement is a real cost, and so is the training time of moving a team off a tool they know.

The view is keyword-mapped. That helps a monthly slide: which money queries now have an AI-shaped answer Semrush modeled. It is a poor source of ChatGPT briefs when the account team talks in comparisons and "should I switch" language nobody typed as a keyword. Buyers do not type keywords into ChatGPT. They ask questions, often long ones, often in the words they used with a salesperson. A modeled keyword graph is a reasonable proxy for demand, but it is a weak proxy for the actual prompt, and the gap shows up the first time you read a real ChatGPT answer next to a Semrush topic row.

We do not cancel the Toolkit to make a point. We pull it monthly. We label it modeled. We write the weekly ChatGPT list somewhere else. Two scoreboards, two labels, one retainer.

How we do topic research for ChatGPT without discarding Semrush

1. Keep the Semrush topic and keyword map. Do not pause it. Rank tracking still pays the bill on Google, and Google is still the majority of the traffic for most of our clients.
2. Write 15 to 25 prompts in sales language. Do not paste the Toolkit export. The prompts should sound like the buyer on the call, not like the keyword column in a spreadsheet.
3. Load them into Promptwatch. Tag by campaign and persona. Set country when the market is not one country, because a French buyer's prompt and a US buyer's prompt can return different answers and different competitors.
4. Read citation analytics and fan-outs. The topic is often "earn a mention on this URL," not "publish our own cluster." If the model already cites a publisher the client cannot displace, the brief is PR, not a new page.
5. If ChatGPT cites a domain Semrush already knows, use Semrush to target it. Promptwatch proves the model still uses it. The two tools answer two different questions, and the combination is stronger than either alone.

How to get cited by ChatGPT is the page edit. Search Central stays the Google rules. GSC generative reports stay for Overviews.

Ahrefs Brand Radar is $199/mo plus an Ahrefs plan. An independent January 2026 check reported 3 ChatGPT mentions where 123 were verified. We keep Ahrefs too. Same split: modeled index versus typed prompts. Ahrefs is excellent for links, and the backlink pivot is the reason to keep it, not the Brand Radar column.

Otterly ($29, Gemini and Claude add-ons, seven-day lag), Peec ($95, three models, Claude Enterprise, extra $35 to $165), and Profound Starter ($99 annual, ChatGPT only) are monitors, not Semrush replacements. They can sit next to Semrush. None of them replaces the keyword graph, and Semrush does not replace the prompt log.

Professional is $245/mo and is the first brand plan with Agent Analytics, the crawl check before a content sprint. The agency plans include Agent Analytics too. Essential has no listed crawler-log allowance. Business is $579/mo. Visitor analytics through a script or GTM attach sessions, which is how a mention becomes a measurable referral rather than a screenshot.

What we will not do

We will not sell the Toolkit as ChatGPT topic research. It is modeled prompts on a keyword graph, and that is a useful object, but it is not the list of questions a buyer asks an assistant. We will not drop Semrush because we added Promptwatch. The two do different work. We will not spawn a URL per modeled variant, because that is how a content program turns into a pile of near-duplicate pages that Google then has to crawl.

FAQ

We already pay $99/domain. Do we still need Promptwatch?

If the SOW includes ChatGPT visibility on a weekly cadence, yes. The Toolkit is the add-on you already have. It is not that ledger.

Are Semrush topics the same as Promptwatch topics?

No. Same English word. Different object. A Semrush topic is a modeled cluster over a keyword graph. A Promptwatch topic is a label on a prompt you typed and the answers that came back.

Can we use Explore to test the workflow?

Yes. Ten ChatGPT prompts. Then Essential if Perplexity or Gemini is in the brief.

What to do this week

1. Keep Semrush. Do not pause rank tracking.
2. Write down whether the AI Toolkit is on the plan. Pull it monthly if it is.
3. Write 15 ChatGPT prompts from sales, not from the Toolkit.
4. Load them into Promptwatch Essential ($95) or Kick-off ($199).
5. Email hello@1001seomedia.com with the Semrush workspace note if you want both scoreboards labeled correctly.