Semrush AI Visibility Toolkit: what to do after setup

After setting up the Semrush AI Visibility Toolkit, work through it in four steps. Use Visibility Overview to find gaps, Brand Performance to see how AI frames you, Prompt Tracking to build a stable baseline, and the AI Search Site Audit to remove blockers. Then turn Topic and Source Opportunities into tickets with owners. The score shows where you stand; the opportunity lists tell you what to do.

This guide covers how to read each report, what to fix first, and when outside help pays for itself.

01What does the toolkit include, and what are its limits?

As of October 2026, the AI Visibility Toolkit costs $99 a month and requires a Semrush subscription.^[1] It includes 25 tracked prompts, one domain for Brand Performance, and AI search checks for up to 100 pages in Site Audit. Extra prompts cost $60 a month per 50, and extra Brand Performance domains $99 each.^[1] Semrush One bundles include the toolkit with higher limits.^[1]

Two limits are worth knowing. Brand Performance uses synthetic prompts generated from your business profile, not real user queries,^[2] so treat it as a model of how AI talks about you. And Semrush does not publish exactly how the AI Visibility Score is calculated in its metrics guide,^[3] so use it for trend and competitor comparison rather than as an absolute grade.

Semrush became part of Adobe in April 2026.^[4] Larger organizations may also be offered Semrush Enterprise AIO, which is sold by demo.^[5]

02How should you read the Visibility Overview?

Start here, but skip past the score. The useful parts are the two opportunity lists. Topic Opportunities are prompts where competitors are visible and you are not. Source Opportunities are external sites citing competitors but not you.^[6]

Export both and sort them before acting:

• Tag each topic by business value: does it match a product you sell, in a market you serve?
• Tag each source by type: review site, directory, publisher, community, partner, or competitor-owned.
• Drop what you cannot or should not chase, such as competitor blogs or topics outside your offer.

What remains is your backlog. The Cited Pages section shows which of your own pages AI already cites,^[6] and those are the pages to keep accurate and current first.

03What does Brand Performance tell you that the score does not?

Brand Performance shows how AI platforms describe you, including sentiment and the narratives that come up, across Google AI Overviews, AI Mode, ChatGPT, Perplexity and Gemini.^[2] Semrush's own guidance makes the point that AI platforms give narratives and summaries rather than ranking pages.^[2]

Look for three things:

1. 1Wrong facts. Old prices, discontinued products, outdated leadership. These go straight onto the fix list.
2. 2Unwanted framing. Being described as "budget" when you sell premium, or as a single-product company when you have a range.
3. 3Negative themes. Repeated complaints. Check whether they reflect real reviews. If so, the fix starts with the product or support team.

For a method of tracing each wrong claim to its source, see how to fix wrong AI information.

04How should you set up Prompt Tracking?

Twenty-five prompts go quickly, so choose them carefully. Pick the prompts closest to revenue: category questions in your main market, comparisons with your top two or three competitors, and factual questions about pricing or availability.

Write each prompt the way a buyer would type it, in full sentences, and in the language of the market. Keep the set fixed for at least a quarter. AI answers vary from run to run: a SparkToro and Gumshoe study found the odds of the same recommendation list appearing twice were under 1 in 100.^[7] Changing prompts every month makes it impossible to tell real movement from noise.

05What should you fix first?

Work from what you control to what you do not. This order usually ships the most in the first month:

1. 1
   Clear Site Audit blockersWeek 1

   Fix crawler access, rendering and indexing issues the AI Search checks flag on key pages.
2. 2
   Correct facts on cited pagesWeek 1–2

   Update the pages AI already cites so the facts it pulls are current and explicit.
3. 3
   Fill high-value topic gapsWeek 2–4

   Create or rewrite pages for topics where competitors appear and you sell the product. Lead with a direct answer.
4. 4
   Work Source OpportunitiesWeek 3 onward

   Claim and complete directory and review profiles, pitch relevant publishers, contribute to community threads honestly.
5. 5
   Re-check monthlyMonthly

   Compare Prompt Tracking and Brand Performance with the previous month, alongside a log of what shipped.

The fourth step usually decides the outcome. In Ahrefs' study of 75,000 brands, branded web mentions correlated with AI Overview visibility at 0.664, versus 0.218 for backlinks.^[8] Source Opportunities are Semrush's map of where those mentions are missing.

06How do you write pages that close topic gaps?

Topic Opportunities often lead teams to publish a new blog post for every gap. That rarely helps. Start by checking whether an existing page should answer the question and simply does not do so clearly.

Pages that get cited tend to share a few traits. They answer the question in the first paragraph, state facts plainly with numbers and dates, use headings that match how people ask, and keep one topic per page. Product and pricing pages matter as much as articles, because many buyer prompts are factual.

Keep the facts consistent everywhere. If your pricing page, a partner listing and an old press release give three different numbers, AI answers may repeat any of them. Before publishing, check the pages Semrush lists as cited sources for the same topic and align them.

07How do you report progress to leadership?

The AI Visibility Score is easy to put on a slide, which is exactly why it gets over-used. Leadership will ask why it moved, and the score alone cannot answer. Pair it with three supporting numbers each month:

• Changes shipped, split by on-site and off-site, from your change log.
• Accuracy: how many of the wrong facts you found are still appearing in Brand Performance.
• Share of voice on priority prompts from Prompt Tracking, against your two closest competitors.

Whatever format you export, add a short note in plain language on what the team did and what it expects next. Keep the same three numbers every month, so the trend becomes the story rather than any single reading. If a number moves sharply, check whether the prompt set, competitor list or engines changed before treating it as a result. Our AI visibility dashboard: what next guide has a 90-day reporting cadence that works with any tool.

08When is it worth bringing in a consultant or operating team?

The toolkit gives you a clear list. Whether you need help depends on whether the list gets shorter. Consider outside help when:

If you cross that line, a partner should work from your Semrush data rather than replace it. Our done-for-you AI visibility guide covers how to scope a pilot.

OwnlyReal works this way: it is tool-agnostic, uses the measurement platform you already have, and ships fixes and authority work inside your approval chain. The enterprise section covers multi-market governance.