Documentation
AI Visibility Features/Tracking a Category

Tracking a Category

View MarkdownAsk ChatGPTAsk ClaudeAsk PerplexityAsk Copilot

Most tracking starts with a subject you own. A category property starts with a market instead. You name the category, describe it in a sentence, and Spyglasses ranks every brand AI names when buyers ask about it. There is no website, no client, and no focal brand.

Product overview: Category Tracking for what it does, who it's for, and how to get started.

What You'll Learn

  • What a category property is and when to reach for one
  • How share of answers differs from share of voice, and why the values don't sum to 100
  • What the Brands list shows and how brands get on it
  • The setup order, and why the report has to run before the brand list is useful
  • How locations give a category a leader per market
  • Which modules a category deliberately does not get, and why
  • The three workflows built on category data

What a Category Property Is

A category is a market you want to watch. "Energy providers for AI data center buildouts." "Project management software for construction firms." "Med spas in the Southeast." You describe what buyers are choosing between, and that description is all Spyglasses needs.

The difference from every other subject type is that nobody is the focal brand. A brand, product, or person property measures one subject against named competitors. A category measures the whole field at once and lets the ranking tell you who the field is.

That makes it the right tool in two situations:

  • You don't have the client yet. A PR or marketing agency can track a vertical before anyone in it has signed, then walk into the pitch with the answers already in hand.
  • You're sizing your own market. An in-house team can watch the category around their product without folding the measurement into their own brand's numbers.

If you already have a brand to track, track the brand. A category runs alongside it rather than replacing it.

What You See

Brands in this category

The main list is called Brands, not Competitors, because there is no "you" for them to compete with. Every brand AI names in a tracked answer lands here, ranked by share of answers.

Brands arrive two ways. Discovery finds them in the answers as they run, and you can add or remove them by hand. Discovery is additive: it only ever adds names that aren't on your list yet. Deactivate a brand and it stays on the list, switched off, so discovery won't bring it back. A brand you delete outright can be found again by a later report.

Share of answers

Share of answers is the percentage of completed answers a brand appears in. If 1,280 answers completed in the window and a brand was named in 157 of them, its share is 12.2%.

The values do not add up to 100. One answer usually names several brands, and plenty of answers name none at all. A category ranking is not a pie chart. Reading it as one will make every brand look smaller than it is.

This is what separates share of answers from share of voice. Share of voice divides a fixed pool of mentions among a brand and the competitors you named, so it does sum. Share of answers has no fixed pool and no named set; it asks one question of every brand independently, which is "how often does AI name you here?"

Movers

Each brand carries a 7-day change. That is the signal to watch on a category, more than the absolute ranking. The leader rarely moves week to week. The brand that climbed four points did something, or somebody wrote something, and that's the thread worth pulling.

Markets and leaders

Once a category has locations, each market gets its own ranking and its own leader. Those leaders often differ from the category-wide leader and from each other, which is usually the most useful thing on the page for anyone selling regionally.

Key messages

Key messages work on a category the way they do on a brand, with the subject shifted. Instead of "which claims does AI repeat about us," a category asks "which claims does AI repeat about this market." Track the framings that matter and you can see whose story the category is being described in.

Citations and publishers

Every tracked answer carries its sources, so the full citation stack is available: Citation Intelligence, Coverage Groups, and Publisher Lookup. For a PR team this is often the point of the exercise. The publishers AI reads for a category are the publishers worth pitching, and you can find them before you have a client in it.

Setting One Up

The order matters, because each step feeds the next.

  1. Name it. Give the category a name you'd recognise in a list of properties.
  2. Describe it. One sentence covering what buyers are choosing between and who those buyers are. This is what the prompt generator reads, so a vague sentence produces vague prompts.
  3. Generate the prompts. Spyglasses writes the questions buyers put to AI for this market. Review them, cut the ones that miss, and add your own.
  4. Run the report. The first AI Visibility Report is what finds the brands. Until it completes, the Brands list is empty, because nothing has been asked yet.
  5. Review the brands. Remove anything that isn't really in the market. Removals stick.
  6. Track daily. Put the prompts on Daily Prompt Tracking and the ranking starts building history.

Fewer than five prompts blocks the report. A category has no site to fall back on, so the prompts are the entire instrument. The report will not start until at least five are active.

Locations for a Category

A category can parent LOCATION sub-properties exactly as a brand or product can. Each location inherits the category's description and prompts, and runs them with that market as context.

What comes back per market is a full ranking, not just a number: the brands AI names in that city or region, in that market's order, with that market's leader at the top. Billing follows the usual rule, so a tracked prompt on a category with N locations consumes N + 1 prompt runs per day. See the Locations dashboard.

The Category AI Visibility Report

A category report leads with who AI names. The ranking is the headline, followed by the movers, the markets, the key messages and the citation stack.

It has no consistency section, no site readiness section and no "your brand vs. theirs" framing, because none of those have a subject to attach to. What it does have is the thing a brand report can't give you, which is the whole field in one view.

What a Category Doesn't Have

Each of these is withheld for the same reason. They all assume a website belonging to the subject, and a category has none.

  • Site audit and AI readiness need pages to crawl and score.
  • Brand consistency asks whether AI describes one subject the same way every time. A category has no such subject.
  • Grounding search rankings and SEO opportunities compare a focal domain's rankings against competitors'. There is no focal domain.
  • Earned-media opportunities are built from "publishers citing your competitors but not you," which needs a "you."
  • AI Traffic Analytics, pages, sitemap import and robots.txt analysis all measure a site you control.

Everything that measures the market rather than a site is available: prompts and tags, key messages, citations and publishers, coverage groups, locations, projects, and the daily runs behind all of it.

The Workflows

Three workflows are built on category data:

  • W29 Map a category. Stand up the property, generate the prompts, run the first report and read the field.
  • W30 Turn a category map into prospects. Work the ranking into a pitch list, with the publishers AI already reads for the market attached.
  • W31 Watch the narrative in a category. Track key messages and movers over time to see whose framing the market is being described in.

Pitfalls

  • The report won't start. You have fewer than five active prompts. A category has no other instrument, so this floor is enforced.
  • A brand you deleted came back. It shouldn't. Discovery only adds names it has never seen, so manual removals stick. If a name reappears, it is a near-duplicate spelling rather than the same record; merge the two.
  • The shares look low. They are shares of answers, not slices of a pie. A 12% leader in a broad category is a strong position. Compare brands against each other and against last week, not against 100.
  • The prompts read like brand prompts. Keep your own brand and every other brand out of them. A category panel should describe the problem and the buyer; the whole point is to see who AI volunteers. See Choosing Prompts.
  • One market's leader surprises you. That is usually correct rather than a bug. Regional leaders diverge from national ones more than most people expect.

Previous

Tracking Products

Next

Tracking Awareness

On this page

How-to guides for setting up Spyglasses to track your AI traffic