I'm genuinely surprised by how Microsoft Clarity keeps shipping new AI visibility features for SEO professionals.
The latest addition to that suite is a beta feature called Topic Insights. I spent a little time around it, and I really appreciate that the Microsoft team is constantly delivering something that is useful.
The pitch on the landing page sums up what Topic Insights does nicely: "Measure how your content shapes AI answers, compare visibility across key topics, see where competitors are cited, and uncover opportunities to strengthen your authority."
Getting started with Topic insights
To get started, Clarity gives you three AI-generated topic suggestions to jump-start a report.
In my case, "Technical SEO strategies," "Search analytics reporting," and "Structured data implementation" — each pre-populated with example prompts that a real user might type into an AI assistant. You can also skip the suggestions entirely and build a topic from scratch, which is really the more useful path once you know what you're trying to track.
For the first report I built, I just went with the AI suggestions. As I didn’t really review the prompts, the report is not that useful, so I ended starting from scratch again.
One thing worth flagging: once a topic is set up, Clarity keeps generating reports on a recurring basis automatically. Each one is logged with the date, the AI model used, and a link back to the prompt set behind it.
Currently, that cadence seems to be fixed as a weekly update. There's no setting to speed it up, slow it down, or trigger an off-cycle report on demand. For a beta, that's a fair trade-off, but it's the one piece of control I'd want added before relying on this for anything time-sensitive.
Setting up your own topics and prompts
Before generating a report, you can create a topic around any theme relevant to your business, then write your own 10–15 prompts underneath it, mirroring the actual questions your audience is likely asking an AI assistant.
Clarity runs those prompts against a live model (it's showing ChatGPT 5.3 at the time of writing), then scores your content's presence in the resulting answers.
In practice, this means you can build out a whole map of the topics that matter to your business — one for each core service, product line, or pillar page — and get a dedicated report for each, rather than a single generic visibility score.
Once a report runs, you land on a per-topic dashboard showing your citation performance, your share of authority relative to Google, your competitors, and everyone else pulling citations in that topic — plus two genuinely useful panels: a breakdown of content gaps (subtopics where others are getting cited and you aren't), and a view of which of your own pages are actively feeding into AI answers.
The layout is clean enough to screenshot or export straight into a stakeholder update, without needing to translate it first. If you've ever struggled to explain "AI visibility" to a client or leadership team, having a dashboard that visually maps citation share, content contribution, and competitor presence in one place does a lot of that explaining for you.
In the report, you can open a pop-up window to View User Prompt and see when you are being cited. Sadly, there is no information about which pages are cited for though.
Defining your own competitors — and seeing everyone else too
In the dashboard, there's an Edit competitors screen where you list the domains you want benchmarked against your own.
This is one of the more thoughtful parts of the feature: rather than just comparing you against a fixed list, Clarity still surfaces every domain that shows up in the AI answers, including sources you never explicitly added as a competitor.
So you get your defined competitive set for direct comparison, plus a wider "everyone else" view showing who else is quietly picking up citations in your space.
The prompt tracker that I'm using elsewhere just compares you against a narrow, pre-set comparison group. So having a controllable competitor list alongside full visibility into untracked domains gives a noticeably more complete picture than most AI visibility trackers.
Top content opportunities
This is the part of the report where things get a bit more subjective. Unlike the citation and authority numbers above, which are fairly numerical counts, the content opportunities section is Clarity's own read on why other sources are getting cited.
Expanding each "Top content opportunities" card breaks down how your competitors are covering a topic and what are the content angles that you are missing. It's effectively a mini content brief for closing the gap, rather than just a list of competitor names.
The downside is that this breakdown is AI-generated, so you should treat it with a grain of salt.
It's a useful starting point for a content brief, but this is exactly where a human-in-the-loop matters: treat it as a first draft to validate, not a finished analysis to act on directly.
It's also worth noting that this section only surfaces the domain, not the actual cited page. If a domain shows up across several content opportunity cards, there's no way to tell whether it's the same page getting cited repeatedly or several different pages on that domain, which makes it harder to actually go compare your content against the specific page that's winning the citation.
Your top content to AI responses
This is a separate panel that shows which of your own pages are actually getting pulled into AI answers, and why.
For my structured data topic, /event-schema-multiple-dates/ and /brand-schema-markup/ were cited for schema-focused explanations that distinguish related markup types and map choices to specific technical scenarios, while /no-referring-sitemaps-detected/ was picked up for diagnosing why certain indexing or sitemap states show up in Search Console.
That's genuinely useful for seeing which of your pages are earning their keep in AI answers — but it comes with the same caveats as the opportunities section. The attribution is tied to the topic as a whole, not the exact prompt that triggered the citation, so there's no way to tell which specific question your page answered, or how many of your 10–15 prompts actually referenced it.
And as with the rest of the report, you still can't see the underlying AI answer itself — just Clarity's summary of why the page was cited, not the actual sentence or context it appeared in.
Where it's still limited
That said, no beta is perfect, and a few things stood out to me as I dug deeper into the reports.
No visibility into the actual AI answer
You get the metrics, such as citation rate, share of authority, content contribution, but not the underlying generated response itself. That makes it harder to judge how your content was used or framed, only that it was cited.
It's also not possible to see the exact wording or context the AI used around your citation, which limits how much you can fine-tune content based on the report alone.
The top content opportunities give you some direction , but it’s also limited as we only know the domain but the actual cited page.
Single-model tracking
Reports are currently generated against ChatGPT only, not a multi-model view. Given how differently ChatGPT, Gemini, Claude, and Perplexity can cite and summarize the same content, this is a meaningful blind spot if your audience isn't exclusively on one assistant.
Competitors are shared across all topics.
The competitor list you define isn't topic-specific — it applies globally across every topic you track. That's a real limitation if your competitive set genuinely differs by subject, and it means you're forced into a one-size-fits-all comparison group.
From my experience, if a company has multiple product offerings, the competitor list can be vastly different for each offering. So it would be helpful to have a topic-level competitor list, rather than one global set applied everywhere.
A tight report allowance
You're capped at 10 reports per project per week, and each topic only supports 10–15 prompts. For a beta, that's a pretty small sandbox, it's barely enough to cover a handful of core topics with any real prompt depth, let alone map out a full content strategy or run comparisons across multiple audience segments.
If you're managing more than a couple of pillar topics, you'll bump into that ceiling fast.
No visualization over time
The History section logs each report as a flat card — date, competitive share, citation rate, and the AI model used — but there's no chart or trendline tying them together.
Right now, with only one report on record, that's not a big deal. But as history builds up week over week, comparing performance across dates means clicking into and mentally tracking numbers across separate cards rather than seeing a trend at a glance.
For a feature explicitly built around tracking change over time, the lack of any visualization here feels like a missed opportunity and one I'd expect to become more noticeable as the report history grows.
Let’s hope a more intuitive trendline view will be added as this comes out of beta.
Early take
As a beta, Topic Insights is a very solid free AI visibility tool. It covers a lot of what paid AI visibility tools charge for: citation tracking, share of authority, competitor benchmarking, content gaps.
While Google Search Console's Generative AI performance report still only shows impressions, with no citation detail or context. Even in beta, Clarity is already ahead of that.
If you're already running Clarity and have access to AI Visibility, it's worth setting up at least one topic around a core content pillar just to see where you stand — and to have something concrete to show the next time someone asks how your content is performing in AI search.