The DataForSEO LLM Mentions API and the Verified Tools AI visibility checker answer different questions. DataForSEO's index shows where your brand already appears inside a fixed pool of AI prompts it has run; the Verified Tools checker asks your own questions live and reads the answers. On 2026-09-12 we ran both on the same two domains. The index scored one of them at zero — a domain where live checks two days earlier had measured 35 real citations.
Neither tool is wrong. They are different instruments, and if you buy one expecting the other's answer, you will misread your AI visibility. Here is the experiment, with every number dated.
The two instruments
An LLM mentions index (we used DataForSEO's LLM Mentions API, the same source the open-source SEO platforms build on) is retrospective. DataForSEO runs a very large pool of prompts against AI platforms on its own schedule — the product page cites ~359 million prompts — and stores where each brand and domain appeared. You query the archive: "across everything already asked, where did my domain show up?" Pricing is $0.10 per request plus $0.001 per row; our whole two-domain probe cost $0.91.
A live check is prospective. You choose the questions, ask the engines right now, and parse the answers — citations, brand mentions, and who got recommended instead of you. Verified Tools runs its visibility checks this way: every report is a fresh set of prompts sent to ChatGPT, Claude, Perplexity, Gemini, and Google's AI Overview at request time.
What the index saw (2026-09-12)
verifiedtools.info — 2 appearances on ChatGPT, 0 on Google AI Overviews.
Both appearances came from productivity questions ("what is the best free online organizer?", "are there any free organizing apps?"), and both were in the web results ChatGPT consulted, not in the answer text itself. The answers recommended other products. That distinction — surfaced in retrieval, absent from the answer — matches what our own logs show from the other side: ChatGPT's crawler fetched our pages 101 times in the week ending 2026-09-10, while answer citations on our 53-prompt tracked set that week were zero. Being read is not being cited.
paircare.love — 0 mentions, every platform.
This is the number that would mislead you. A live scan of the same domain on 2026-09-10 — 30 fixed prompts across 5 engines — measured 35 citations, including 5 from ChatGPT and 1 inside Google's AI Overview. The domain is demonstrably visible in AI answers. The index says zero.
Why a cited domain scores zero in the index
Sampling. The index can only report on prompts that exist in its pool, and a pool built for breadth concentrates on mass-market questions — best CRM, best to-do app, best email tool. paircare.love competes in a narrow consumer-health niche ("cycle tracking for couples"), and the questions it wins simply are not in the archive. Zero mentions there means "we never asked the questions you win," not "AI never recommends you."
The reverse limitation applies to live checking: it only sees the questions you thought to ask. A frozen prompt set is deliberately narrow — that is what makes week-over-week comparison honest — so it will not tell you about an unexpected productivity query where your domain surfaced once in March.
When each one is the right tool
| You want to know | Use |
|---|---|
| "Across the mass-market question space, does my brand appear at all — and who gets cited where I don't?" | DataForSEO's mentions index (~$0.45/brand as measured) |
| "Do AI engines recommend me for the specific questions my customers ask?" | A live checker like Verified Tools |
| "Did last month's content changes move my citations?" | A live check with a fixed prompt set — an index pool changes underneath you, so before/after deltas aren't attributable |
| "Quick screening of a mass-market brand before deeper work" | Index first, live check to go deep |
| Any niche or non-US/non-English category | Live check — the index's coverage there is too thin to interpret zero |
One more structural note: the index we tested reports ChatGPT data for the US/English market only, and its archive refreshes on a monthly cycle. Live checks have neither constraint — you pay per question instead.
What we did with this result
We decided not to adopt the mentions index into our own tracking. Our measurement program is built on a frozen prompt set scanned weekly, because the point is attribution — knowing whether a specific page change produced a citation a few weeks later. An archive whose prompt pool we don't control can't support that, and for the niche verticals we work in, its zeros are unreadable. For screening a mass-market brand cheaply, it remains a reasonable first pass.
All figures on this page are from our own API calls and scan logs on the dates shown. We paid standard list price for the API and have no affiliation with DataForSEO beyond being a customer.