Competitive intelligence tools for AI search: how we run it for clients

Competitive intelligence for AI search is, in our experience, the part of a program clients enjoy most and misuse most. Seeing a rival named in every answer is memorable. Turning it into a plan takes more than a screenshot.

The questions we answer

For each client, we want to know five things. Who is named instead of us, and on which prompts? Which sources do the engines cite for those prompts? How do the answers describe the rival compared with us? Which competitors are we not tracking that we should be? And, where ad formats exist, who is paying to appear?

Notice that only the first is a visibility number. The other four are diagnosis.

How we set it up

We fix a competitor set at the start, so share of voice stays comparable across months. We add rivals deliberately, not casually, because every addition shifts the denominator. We track by topic, since a client can lead overall and still lose an entire category. And we tag prompt types: organic, brand and competitor comparison, since comparison prompts often decide shortlists.

What we run it on

We use Promptwatch for client programs, and its competitive features are a big reason. Competitor visibility, share of voice and position sit next to heatmaps that compare any two dimensions, such as brands against topics or models. Auto-discovered competitors surface rivals we had not listed, and brand merging keeps duplicates from skewing the numbers.

The citation tools do the diagnosis. Page Tracker follows specific URLs, ours or a competitor's. Citation analytics shows pages, domains, Reddit, YouTube and LinkedIn citations, plus offsite mentions where a brand is named without a link. Citation Trends classifies each citation by content type and source type, so we can see whether a rival wins through reviews, comparison pages or forum threads.

Ads Radar, which starts on Business and is on every agency plan, captures sponsored ads inside ChatGPT Search, Google AI Overviews and Google AI Mode answers to tracked prompts, with the advertiser, copy, landing page and triggering prompt. It does not cover Perplexity, and we tell clients that.

To act on all of this, content gap analysis shows what the rival covers that the client does not, and the Action Board orders the tasks. Plans as of late October 2026: Explore free (10 prompts, ChatGPT only), Essential $95 a month, Professional $245, Business $579, with agency plans at $199, $399 and $799.

Other tools we have looked at

AthenaHQ offers unlimited competitors on Starter, with credit-based usage. Scrunch AI offers competitor benchmarking with site audits. Profound reports citation share inside an enterprise platform. Evertune samples each prompt many times across 11 models, with no published pricing. Ahrefs and Semrush add competitor comparisons to their suites. We choose based on whether the tool can explain a rival's lead.

Our review routine

Compare share of voice by topic against the fixed competitor set.
Pick the topic where a rival gained most.
Open citations for it: which sources and page types win?
Check whether the client has a page that answers the question at all.
Plan one fix, ship it, and re-measure the same prompts.

Realistic expectations

Competitive intelligence tells you where a gap is. It does not close it, and no tool can guarantee that you overtake a rival in an AI answer. Answers vary by run and country, so we read direction across weeks.

If you are starting, track your three closest competitors on 30 buyer prompts and read the citations before you write anything.

A note on client reporting

We keep the competitor slide simple: one chart of share of voice by topic, one list of the sources winning for the biggest loss, and one proposed fix. Clients can absorb that in a meeting. The full heatmap lives in an appendix for the people who want to dig in.