An AI visibility dashboard shows you whether AI platforms citation and propose your brand.
We’ll assistance you build one that puts your prompt-level AI data from Semrush next to your page-level Google data from Search Console, so you can see whether the pages AI cites are additionally earning clicks.
On our test site, the archetypal refresh surfaced a guide that AI answers cited for 50 prompts and that earned zero Google clicks.
To build your own brand visibility dashboard for AI search, you’ll need Semrush's AI Visibility Toolkit, Google Search Console (GSC), Google Sheets, and a paid Claude plan. The Sheet holds your data. Claude turns it into a live, interactive dashboard. It took me concerning 90 minutes to build out this dashboard the archetypal period and it takes 20 minutes to update all week alongside caller data.
What this AI visibility dashboard volition show
This dashboard shows two measurements flank by side: how apparent your brand is in AI answers (from Semrush) and how your pages execute in Google (from GSC). Specifically, it tracks:
- AI visibility rate: the portion of your tracked prompts anywhere your brand appears
- Mentions and cited pages from Semrush's immediate tracking
- Generative AI impressions from GSC's Generative AI achievement report
- Clicks, impressions, click-through charge (CTR), and average stance from GSC
- A catalog of SEO and answer motor optimization (AEO) opportunities according to your genuine data
There are two constraints to be conscious of before you begin:
- The dashboard runs as a Claude relic that pulls live data from your Google Sheet through the Google Drive connector. The connector itself is accessible to everyone, but artifacts that call connectors necessitate a paid Claude scheme (Pro, Max, Team, or Enterprise).
- You can't portion it by community link. On Team and Enterprise plans, you can portion it inner your organization, and all spectator sees data lone if their Google document can open your Sheet and their Claude document has the Drive connector. On Pro and Max, it stays personal to you.
This is a operating dashboard for whoever owns reporting, not a nexus you dispatch to leadership. For stakeholder reporting, you can ask Claude to rotate the dashboard’s study into an HTML document they can open in any browser equal without the connections.
Step 1: Decide what you desire to measure
Before you open any tool, decide what you desire to measure:
- Your mark domain, affirmative brand-name variations AI power use
- Which AI platforms you desire to track (Semrush's Prompt Tracking supports ChatGPT, Google AI Mode, and Gemini)
- Location and language
- A stable set of prompts tied to how buyers really conversation to AI
- Your desired reporting period, specified as the latest 28 days versus the former 28 days
Step 2: Set up AI immediate tracking in Semrush
To set up AI immediate tracking, create a run in the Prompt Tracking tool, which checks a fixed set of prompts against the chosen AI phase all day and records whether your brand appears.
Choose and add your prompts
Add prompts that mirror how your buyers conversation to AI. If you can, quarry client assistance tickets, revenue calls, and forum posts. You can nourish transcripts and content posts into Claude to get improved immediate options according to the patterns it sees.
Spread your prompts throughout the complete journey fairly than clustering on brand terms. I used 30 prompts, six in all of five groups:
- Customer problems: The issues your possible customers are dealing alongside (e.g., "I was fair named administrator of my dad's asset and have no idea anywhere to start. What do I do first?")
- Product or category discovery: When your customers don’t cognize your merchandise or assistance exists yet, and they’re trying to discover a solution; e.g., "Is there application that helps executors resolve an estate?"
- Comparisons: Comparing your brand to a competitor’s brand; e.g., "Which is improved for [your ideal client or one particular issue you assistance solve], [your brand] or [competitor]?"
- Product selection: Researching merchandise or assistance options beyond straightforward comparison; e.g., "What are the finest tools for an asset without a will?"
- Branded questions: Questions immediately concerning your brand; e.g., "Is [brand] legit?"
Start alongside about 30 prompts. Monitored daily, that's approximately 200 observations a week, adequate for a stable site-wide visibility rate. The admission AI Visibility Toolkit scheme covers 25 prompts, so 30 method a higher tier or a immediate add-on.
Group-level numbers volition swing. With six prompts per group, one immediate changing moves that group's charge by 17 points, so treat collection changes as leads to investigate. If stakeholders desire dependable group-level reporting, develop all collection to 15 or additional prompts complete time, and lone add prompts that mirror genuine buyer language.
Your tagged prompts volition nearly continually display your brand, which lifts your general rate. Report on all prompts and non-branded only. The non-branded figure tells you whether AI recommends you to group who don't cognize you yet.
Keep the identical center prompts stable for at smallest a duration formerly tracking starts. Every trend row in your dashboard depends on the immediate set staying constant. If you toggle prompts mid-stream, you interrupt the reporting.
Organize your prompts alongside tags
Tag all immediate alongside the collection it belongs to, so you can difference visibility throughout meaningful segments alternatively of one blended score.
In your campaign, open the "Tags" tab and click "Manage tags" to create and allocate them. I used one tag for all of my five immediate groups. The screenshot below is from my Google AI Mode campaign, which tracks a smaller set of prompts throughout the identical five groups. Semrush additionally shows its own topic groupings in this view, and you can disregard those for the dashboard.
Step 3: Export your Semrush AI visibility data
You’ll need to export two datasets from the AI Visibility Toolkit into Google Sheets: prompt-level achievement from the tracking run you set up in stage 2, and the pages AI platforms cite.
Neither dataset is accessible through the Semrush API yet, so the data gets to your Sheet by export. You can agenda those exports via email (more on that below).
Export prompt-level performance
Your immediate tracking run opens in a Position Tracking perspective alongside "AI Visibility" toggled on. Once the data is populated (it may obtain a few hours), scroll downward to the “Top Prompts” division and click “View all X prompts.” Then, click “Export” in the top-right border and choose Google Sheets.
You should see one row per prompt, alongside a pillar collection for all day showing your brand's stance in the answer, the answer type, and the landing leaf cited. A stance of "-" method your brand didn't appear that day. You'll reshape this in the expert Sheet step.
Export cited pages and sources
In Visibility Overview, use the phase selector to choice the identical AI phase your tracking run follows, so your citation data and immediate data depict the identical platform. Then scroll downward to the “Topics & Sources” section. You’re going to export the Google Sheets data for the two the “Cited Sources” and “Cited Pages” tabs.
Visibility Overview holds months of former equal if you created your immediate tracking run yesterday, since Semrush builds it from its own ongoing inspection of the AI answer space. My test location had a mentions trend going rear six months on day one.
What to do if export is unavailable
If the export clasp is grayed out or missing, inspect your admission to the campaign. If person alternatively owns it, ask them to export the reports for you or to alter your permissions. Nothing alternatively in this workflow needs owner rights.
Reach out to Semrush support if you encounter any difficulties here.
Step 4: Export Google Search Console data
GSC data shows what your pages really acquire in Google, next to what AI answers say concerning you. Which study you use depends on whether the Generative AI achievement study has data for your property.
If your Generative AI study has data
Open your asset in GSC, go to "Performance," and open the "Generative AI" report. Export it twice. Once alongside the date range set to the final 3 months, for the regular trend. Once set to the final 28 days, for the pages table, so it matches your norm GSC window.
The export is a ZIP of distinct one-dimension tables (daily impressions, pages, countries, devices). Impressions are currently the lone metric available.
There is no prompt-to-click attribution; Google currently does not provision you queries, clicks, or CTR data. The pages array is aggregated throughout the entire date range fairly than damaged out by day.
If your Generative AI study is empty
The Generative AI study is accessible to all property, but it shows data lone formerly your pages get adequate impressions in AI Overviews and AI Mode. If yours is bare or thin, use the norm "Search results" achievement study instead. Export page-level clicks, impressions, CTR, and average stance for two matched 28-day windows.
End all export opening 2-3 days before today, since the newest GSC data is incomplete and volition undercount. Keep the windows the identical dimension so duration comparisons are consistent.
Note that this is general Google Search performance, not AI-specific visibility, so tag it that way in your dashboard.
Step 5: Organize everything in one expert Google Sheet
Put all export into one Google Sheet alongside a fixed tab structure. Mine ended up alongside seven data tabs, all holding one dataset:
- semrush_prompts: prompt-level results reshaped to one row per immediate per day, alongside columns for date, prompt, platform, tag, position, and cited landing page
- semrush_cited_pages: the cited pages export, stamped alongside a snapshot date
- semrush_source_domains and semrush_topics: the Visibility Overview exports, identical stamping (topics is optional)
- gsc_genai_daily and gsc_genai_pages: the regular array from your 3-month Generative AI export and the pages array from your 28-day one
- gsc_pages: norm page-level GSC data, one row set per 28-day window
Here is a immediate you can nourish to Claude alongside alongside all the exports to have it build the panel for you:
On forthcoming weekly updates, I'll nexus the current expert panel (downloaded as .xlsx) affirmative new exports.
Step 6: Turn the Google Sheet into a live Claude artifact
Next, we’ll rotate the Sheet into a dashboard by asking Claude to build a live relic that says the Sheet through your Google Drive connector.
Connect the Google Sheet to Claude
Connect Google Drive in Claude's connector settings, afterward verify Claude can open the expert Sheet before asking for item visual. The connector lets the relic re-read the live Sheet on demand, so it updates all period you add the new data.
Tell Claude what to include
Prompt Claude to build the ocular dashboard connected to your expert sheet. Make certain you substitute [sheet name] alongside your Sheet’s name and [domain] alongside your domain:
The freshness warnings halt you from presenting old data as current, and the "compare, never blend" education keeps data separate. The chance division in that immediate is how the dashboard makes recommendations according to the data.
On your archetypal build, the prompt-tracking charts volition display lone a few days of data. That's expected.
On our test site, the archetypal refresh flagged three things that went direct onto the satisfied plan. Step 8 walks through all one.
Step 7: Test that the dashboard updates correctly
The entire item of the dashboard is to track visibility over time, so you’ll desire to create certain it updates properly. Test the association by changing one figure in the Sheet and confirming it moves in the dashboard.
Edit a clicks value in the “gsc_pages” tab, refresh the artifact, and inspect that the matching KPI cardstock changes. Check the data-freshness stamp shows the current read, afterward undo the edit.
If a division of the dashboard goes bare following a Sheet edit, difference your pillar headers against the first structure. They need to equivalent exactly.
Step 8: Review the dashboard and prioritize actions
Use the dashboard to decide what to activity on next. Here's how we peruse the three findings from our test site's archetypal refresh:
- A leaf AI cites that earns no Google clicks: Our administrator checklist was cited for 50 prompts and earned zero clicks. AI platforms already rely the page, so the activity is on the Google side. Find a keyword it can realistically position for, refresh the leaf about it, and build links to it.
- Competitors cited anywhere you aren't: Our competitor list was led by domains cited for seven prompts each. Open those prompts and peruse what gets cited. That's your catalog of difference and category satisfied to create, and of listicles to get mentioned in.
- A entire immediate collection alongside no presence: We had zero visibility throughout all six customer-problem prompts. One feeble leaf doesn't explain that. It points to missing content, so those six prompts became the concise for new top-of-funnel pieces.
As your former builds, observe for these patterns too:
- A leaf ranking fine in Google but rarely cited by AI normally needs clearer structure: straightforward answers, descriptive headings, and current dates that create passages uncomplicated to lift
- A leaf losing mentions following a satisfied alter is your first alert to assessment the edit
- Topics improving throughout the two AI answers and Google hunt are your evidence of what works, so document what those pages have in common
You’ll additionally see an “Opportunities” section, anywhere Claude reviews the overlap between GSC and Semrush data and gives you suggestions on how to enhance your AI visibility.
For example, a leaf can position fine in Google and motionless never appear in ChatGPT. ChatGPT pulls from its own web index, so a powerful Google ranking doesn't justify a citation. Check whether the leaf answers the tracked immediate directly, and whether the sources ChatGPT does citation for that immediate citation you. If Bing and Copilot matter to you, Semrush's Position Tracking tool tracks Bing rankings, and you can add that data to the dashboard alongside one additional export.
Step 9: Update the dashboard all week
You’ll need to update the Sheet on a fixed weekly regular to keep the data accurate. Each week, preferably on the identical day:
- Export the latest immediate tracking data (or collect it from the scheduled export email)
- Export the latest Cited Pages and Cited Sources tables from Visibility Overview
- Export the latest GSC data, ending the opening 2-3 days prior
- Append the new rows alongside the new snapshot date, afterward bring the updated document into the identical Google Sheet alongside File > Import > Replace spreadsheet. If you import it as a new file, the dashboard can keep study the old one.
- Refresh the relic and verify the freshness stamps updated
- Note any important satisfied changes or campaigns in the Sheet, so next duration you can nexus movements to causes
When you update the sheet, keep these tips in mind:
- Keep the identical headers
- Stamp all Semrush export lot alongside its snapshot date
- Append new rows fairly than replacing old ones so you get difference data
- Normalize URLs consistently throughout tabs
If you don’t normalize URLs, GSC can study several of your pages twice; formerly alongside a trailing slash and formerly without. If your page-level numbers always appearance absurdly low, URL variants are the archetypal item to check.
Build your AI visibility dashboard
A brand visibility dashboard for AI hunt shows you how frequently AI cites and mentions your brand, next to what those identical pages acquire in Google.
Open the AI Visibility Toolkit, create your archetypal tracking campaign, and compose 30 prompts the way your buyers really ask. Then nexus that data alongside your expert panel and build the dashboard in Claude.
And if you need to display the dashboard to stakeholders, ask Claude to rotate the relic into a shareable HTML document alongside the identical data.
