AI internet check vs database lookup compared

Most UK reverse-lookup sites give you one of two things: a database lookup (Range Holder, area code) or a community-reports lookup (user comments, scam scores). Each has a major weakness. The third option — a live AI web search — is what we built into WhoCalledLookup, alongside both classic signals. Here's why we layer all three.

#Database lookups: accurate but shallow

Ofcom range data tells you exactly who Ofcom allocated a block to. But it doesn't tell you anything about behaviour. A Sky-allocated 0345 number used by a legitimate utility looks identical, in the database, to a Sky-allocated 0345 number being abused by a scam call-centre. Useful but limited.

#Community reports: behavioural but noisy

User comments capture real-world behaviour — but they're crowdsourced, easily gamed, and slow to update. A new scam campaign starting Monday won't show up in comment threads until people start complaining mid-week. And a legitimate business may have 3 furious one-star reviews next to 30,000 happy customers who never thought to leave a comment.

#AI internet check: fast and cited

An LLM with web-search access reads dozens of forum threads, scam-database entries, business directory pages and Reddit posts in seconds, summarises the consensus, and cites sources. It catches new patterns faster than community boards because it's reading the original source material — not waiting for one site's editor to write a summary.

#Why we combine all three

WhoCalledLookup shows the Ofcom Range Holder (authoritative source-of-truth for the type of number), runs a live AI internet check with citations (behavioural reputation + recent web reports), and links to the major UK community boards from every result page. Either signal alone is half the picture. Together they're hard to beat.

#What the AI is actually doing under the hood

We use OpenAI's Responses API with the web_search tool, scoped to UK-relevant sources. The model gets the number, in three formats (national, international, no-spaces), and asked to:

1. Find any public mentions of the number on UK forums, scam-reporting sites, business directories and Trustpilot.
2. Summarise the consensus in one paragraph.
3. Score risk 0–10 using a published rubric (volume of complaints, recency, severity).
4. Cite at least three sources with URL + date.

The result is cached for 24 hours to keep AI costs predictable, then re-runs on the next lookup of the same number.

#Bottom line

The future of reverse lookup is multi-signal: official allocation data, community reports, live AI synthesis. We built WhoCalledLookup as the first UK consumer site that surfaces all three on one page, free, with no signup. Use them together — don't trust any single one in isolation.