Four different things
AI visibility reports often mix up four separate events. They sit in a chain, and each step loses people. A report that blurs them can make a client feel great, or panic, for the wrong reason.
| Event | What it means | Where you see it | Does Spotlit measure it? |
|---|---|---|---|
| Mention | The answer names your brand | In the answer text | Yes, in its own simulated answers |
| Citation | The answer links to one of your pages as a source | In the source list | Yes, in its own simulated answers |
| Visit | Someone clicks through to your site | Your web analytics | No |
| Lead | A visitor signs up, books a call or buys | Your CRM or analytics | No |
The first two happen inside the answer. The last two happen on your site. An AI visibility tool can observe the first two. Only your own analytics can show the last two.
Mentions: being named
A mention means the answer says your brand name. If someone asks “best email tool for a small online shop?” and the answer lists Mailchimp, Klaviyo and Acme Email, that's a mention for Acme Email.
Two details make mentions more useful:
- Position. Named first with a reason (“best for small shops”) is stronger than named sixth in a list of ten.
- Branded vs unbranded questions. If the question already contains your name (“Is Acme Email any good?”), you'll almost always be mentioned. That says nothing about discovery. Spotlit leaves the branded question out of the visibility score for this reason.
Citations: being the source
A citation means one of your pages is in the list of sources the answer was built from. It's a different thing from a mention, and the two don't always come together:
| Case | Example | What it tells you |
|---|---|---|
| Mentioned, not cited | Acme Email is named, but the source is a roundup on a tech blog | Third-party pages are carrying you. Protect those listings. |
| Cited, not mentioned | Acme Email's blog post is a source, but the answer recommends other brands | Your content is useful, but it isn't pointing buyers to you. |
| Both | Named, and your pricing page is a source | Strong, usually for questions close to your brand |
| Neither | Not named, not a source | Find which pages were used and whether you can get on them |
For unbranded “best X” questions, citations usually go to review sites, roundups, directories and forums rather than brand sites. Our explainer How AI decides which brands to recommend covers why.
Best email marketing tools 2026
Searched the web · read 3 pages
The best email marketing tools in 2026 include Mailchimp (best overall), Klaviyo (best for e-commerce), ConvertKit (best for creators), ActiveCampaign (best automation), and Brevo (best value).
- 1g2.com
- 2pcmag.com
- 3zapier.com
Spotlit reading
- Acme Email
- Not named
- Rivals named
- 5
- Pages cited
- 3
Why one answer is not a score
Ask an AI tool the same question twice and you can get a different answer. The search step may return different pages, and the model may write the list in a different order. So a single answer is a sample, not a measurement.
We check this in our own pipeline. On our research page we ask the same buyer questions several times and record whether each brand was named and in what position. When we wrote this guide, a brand's position moved between runs more often than whether it was named at all. The page shows the live numbers and its limits: the scans are self-selected, the repeat counts are small, and it measures our pipeline, not every AI tool.
- Report a rate, not a yes or no. “Named in 2 of 3 runs” says more than “named”.
- Don't sell an exact rank. “#2 in AI” can be #4 on the next run. Talk about ranges and trends.
- Confirm a change before you report it. A move seen once may be noise. Re-run the same questions first.
How to read results honestly
- 1
Start with the question
Every number should sit next to the exact question that produced it. Change the wording and you've changed the test, so keep the question set frozen between checks.
- 2
Separate mentions from citations
Ask two questions: were we named, and were our pages used? They point to different work.
- 3
Look at rates over several runs
Use the share of runs that named you, and how that share moves over weeks, not one answer on one day.
- 4
Check the sources
The pages behind an answer explain the result. If competitors are named because a roundup lists them, that roundup is your next target.
- 5
Connect to visits and leads in your own analytics
A visibility report can't tell you about traffic or revenue. Look at referral traffic from AI tools in your analytics and at how leads say they found you. Report those numbers as their own line, not as part of the visibility score.
Common questions
Which matters more, a mention or a citation?
For being recommended, the mention, because that's what the buyer reads. A citation helps when the buyer clicks the source. Track both, and don't add them into one number.
Can an AI visibility tool tell me how many customers AI sent me?
No. It can show what answers say. Visits and leads live in your analytics and CRM. Be wary of any report that turns mentions into revenue without that data.
How many runs are enough?
There is no magic number. More runs give a steadier rate. At minimum, ask each question a few times, and re-check the same questions over weeks before calling something a trend.
See your own mentions and citations side by side, across repeat runs: Scan your site free, then read how the Sources page works.
Found this useful? Share it with your team.
We build Spotlit and write these from how our own scans work, with every outside fact linked to its source.