tools

Reading the AI Visibility Report: Which Products Have Gaps and Why

The AI visibility report, section by section: recommendation rate, who's named instead, sites linked, channels, per-product gaps, per-question results. What to act on first.

F
Furniture Connect TeamNovember 20, 2026
Reading the AI Visibility Report: Which Products Have Gaps and Why

The first report is a lot of information at once, and the temptation is to look at the headline number and feel either good or bad about it. Resist that. The headline is the least actionable part. The report is built so that the useful parts are further down: which products, which questions, which missing facts. Here's how to read it in the order that leads to work.

First, remember what it is

One check is a sample of answers produced at a particular time. Assistants can give different answers on another run and change over time. Treat the figures as observations, not precise measurements. The report combines observed answers with gaps in your catalog and channel records; those gaps suggest useful work, but they are not proof of why an assistant omitted a product. Avoid drawing conclusions from a small score difference or one changed answer.

Section one: how often you're recommended

The brand-level figure: across the shopper questions, how often an assistant put your brand in its answer. Note it, date it, and move on. It's your baseline for comparison after you've fixed things, and it's the number that's most sensitive to sampling noise, so don't read a small move as a result.

The report distinguishes this from being known when asked by name. If you're recognized for named questions and absent from shopper questions, the assistants know you exist and aren't reaching for you unprompted. That's a common starting position. Being known by name doesn't establish enough relevance to earn a recommendation, and the catalog gaps the report lists are one place to look, not the proven cause.

Section two: who's recommended instead

The brands that appear in the answers to your shopper questions. Compare this list to the competitors you entered. The ones you didn't expect are the useful ones, and the ones that appear most often are the pages to open next.

For two or three of them, do the side-by-side from what the AI found on your competitor's page that it didn't find on yours: take a shopper question, open their product and yours, and compare which facts each page gives in text. What each page states is often visible in the specifications table; it won't tell you the whole reason.

Section three: which sites are linked

Where the assistants point. If they're linking to a marketplace listing of your product rather than your own page, that tells you the listing was a source used in that answer, not that your page was unreadable. It's still worth checking what the listing states that your page doesn't (often, structured fields). If they're linking to review sites or comparison articles, those are the sources shaping the answer.

Section four: sales channels

How your products show up through the channels you sell on. Read this alongside section three. A product that's listed or ready on one channel and not another has the same record behind it, so the channel's listing is one place to look.

Section five: per-product scores and catalog gaps

This is the section that turns into work. Each tracked product has a score and a list of gaps in its catalog record, such as description, dimensions, materials or images.

Read it across products, not one at a time. If "dimensions" is a gap on twelve of twenty products, that's one attribute to fill across the range, not twelve page fixes. If "images" is a gap on the whole recent launch, that's a Studio job for that range. The pattern across products is the plan; a single product's gaps are a to-do.

Then sort by the question "does fixing this move several products at once?", which is the same test as in why fixing one product page doesn't fix the range. Range-wide gaps first.

Section six: per-question breakdown

For each tracked question, what came back: whether you appeared, who did, and for named questions, what the assistant said about your product. Two things to look for.

Shopper questions you don't appear for. For pet-related questions, check the fabric, care and durability information on the record. For small-space questions, check dimensions and variant information. Treat these as places to look, not proof of why you were left out.

Named questions where the assistant got a fact wrong. These are more urgent than gaps, because a wrong fact travels. Check whether your page says something vague the assistant filled in, or whether the default variant's value is being reported as the product's.

What to write down

From one report, you should leave with three short lists.

  1. Attributes missing across many products, in order of how many they affect.
  2. Products with wrong facts in named answers, and what to check.
  3. Two or three competitor pages to study.

That's enough to start closing the gaps. Re-run the check after the first round and compare the direction, not the decimals.

What not to do

Don't patch pages by hand from the report if the same correction belongs in the source record. Don't chase the brand-level figure week to week. Track the questions you'll act on, and review the displayed credit estimate before each run you start. Scheduled checks run without one, so set the schedule to match.

For how the check is set up and run, see how to run an AI visibility check on your range. For the whole argument, see being found by AI.

Ready to show and sell?

Give every team the product data, imagery and assets they need to launch faster, quote better and turn more of the catalog into revenue.

No slides. We stage one of your products live, at a time you pick.

Talk to salesOr see pricing