1001 SEO Media
All posts
By 1001 SEO MediaGEOlocalSemrushAI searchprocess

Semrush AI Visibility Toolkit: Local Search Tracking and GEO (2026)

When a client already on Semrush asks for local AI tracking: what we set up in Semrush, where we stop, and what we run on Promptwatch for city and storefront GEO.

A brief we hear often goes something like this: "We already pay for Semrush. Can't we just track local AI answers in there?" It's a reasonable ask. Nobody wants another login, and Semrush does sell an AI Visibility Toolkit.

Our answer is that Semrush can do part of it, and we use that part. Then we stop at a specific point and run the local AI program on Promptwatch. This post explains where that line sits and why. Semrush details come from its llms.txt, knowledge base and pricing pages, which we read again on 5 November 2026.

First, we check what the client actually owns

"Semrush" can mean three different purchases, and they don't overlap much.

The AI Visibility Toolkit measures AI answers. Semrush lists it at $99 a month (the pricing page shows "$99/mo per domain, billed annually"), with 25 prompts in Prompt Tracking and no free trial. The Local Toolkit is a separate product for Google Search and Maps, priced per location: Local Base $30 and Local Pro $60 a month, labelled billed annually on the pricing page. Semrush One bundles the SEO suite with AI visibility and 50, 100 or 200 prompts depending on tier.

Many local clients have the Local Toolkit and nothing on the AI side. That's fine. It tells us what's already being measured.

What we set up in Semrush

If the client runs Map Rank Tracker, we leave it running. It's a geo-grid for Google Maps rankings, one credit per pin per keyword, and it's the report most local owners already know. We report it as Maps, under its own heading.

If they use the GBP AI Agent for profile posts, review replies and photos, we keep it on but make sure someone at the client reads what goes out. Automated replies with a wrong opening time are a bad look.

If they have the AI Visibility Toolkit, we point Brand Performance at the client's main city. It's tied to one location and language chosen at setup, from 68,000+ options that include cities, and it updates weekly. That gives the client a familiar weekly read inside the tool they already open. Each extra location costs another $99 a month, so we rarely go past one.

We also borrow Prompt Research. Its volume, difficulty and intent data is a decent sanity check on the prompt wording we plan to track elsewhere, although topic volume only covers 15 countries. And we run the AI Search checks in Site Audit (up to 100 pages) to catch AI crawler blockers early.

Where we stop

We don't build a city program on Semrush's Prompt Tracking, for reasons we can point to in their own docs.

Location is set per campaign, and the knowledge base lists coverage as 220+ countries and territories. It doesn't say a campaign can target a city. If we can't confirm the location, we can't promise the client a city report.

Each campaign runs one engine: ChatGPT Search, Google AI Mode or Gemini. Semrush's blog says the same prompts on two engines count twice against the limit, and that Claude and Perplexity aren't supported in Prompt Tracking. Results are desktop only. Share of Voice isn't available in Prompt Tracking either, and that's usually the first number a client asks about.

Then there's what happens after the answer. The Semrush pages we read don't describe AI crawler logs beyond the Site Audit blocker checks, or any tracking of AI-referred visits and conversions. Without those we can tell a client they dropped in a city, but not why, and not what it cost them.

What we run on Promptwatch instead

We run client programs on Promptwatch. Every monitor carries its own location, language and models, and runs daily.

For service-area towns we use city monitors. Those are on the Business brand plan and on every agency plan. For storefronts we use coordinate targeting: a pin with a radius from 20 m to 5 km, so ChatGPT, Google AI Overviews and Google AI Mode answer as if asked from outside the door. Other models on that monitor use its country, state or city instead. Coordinates are on Business, Agency Scale and Enterprise, and the pin can't be moved later, so we place it carefully. Our setup sequence for new local clients covers how we decide between the two.

Next we connect crawler logs. Agent Analytics verifies AI bots against each operator's IP ranges, so a fake GPTBot doesn't count, and it shows a citation rate per page. When a city drops, that tells us whether the location page wasn't fetched or was fetched and not cited. Visitor analytics goes in through the script, the GTM template or the WordPress plugin. Since ChatGPT stopped sending a referrer on 8 October 2026, we switch on the toggle that counts utm_source=chatgpt.com visits as ChatGPT.

The Action Board then gives us a prioritized list each week from the Discovery agent. If the client is on Business, the Content Agent can draft location pages against the gaps, and we approve every local fact before anything publishes.

How we report both without muddling them

Two sections, never blended. Maps rankings from Semrush go in one. AI answers from Promptwatch go in the other, by city or storefront, by engine. We don't add the scores together, and we don't tell clients that a better Maps position will lift their AI answers. No engine documents that link, so we measure each on its own terms.

When we tell a client Semrush is enough

Sometimes it is. A single location whose customers come through Google Maps, where the owner wants Maps rankings and an occasional look at what ChatGPT says, doesn't need a second platform. Brand Performance on their city and the Local Toolkit cover that, and we say so.

Once there are several cities, storefronts competing street by street, or a question about whether AI answers bring in customers, we move the AI side to Promptwatch. If you want to see where a location stands before deciding, run the free AI Visibility Report and send the result to hello@1001seomedia.com.