How ThriveTalk verifies every therapist profile

Every therapist profile on this site goes through a multi-layer verification process that combines AI-assisted research, automated data collection, proprietary cross-checking tools, and independent fact-checking against primary government sources — and no profile is ever published without a human editor reviewing and approving it first.

This page explains that process in full, because we think you deserve to know exactly where our information comes from.

Before any profile appears on ThriveTalk, it passes through four layers:

1. Research. We gather information about each therapist from their own website, professional directories, the federal NPI registry, and state licensing boards. Anything we can’t trace to a source is left out — our systems are built so that unsupported claims stay blank rather than getting filled in with guesses.
2. Independent fact-checking. A separate verification pass — deliberately walled off from the original research — re-checks every material claim against primary sources: the issuing state board’s own license lookup, the national NPI registry, and federal disciplinary and exclusion databases. It scores its confidence in the profile from 0 to 100.
3. Human editorial review. A member of our team personally reviews every profile — the facts, the sources, the photo, the wording — before deciding whether it publishes. Nothing goes live automatically. Ever.
4. Ongoing re-verification. Published profiles carry the date they were last verified, and we re-run the full research and fact-checking cycle on a recurring basis so pages stay current instead of quietly going stale.

The rest of this page walks through each layer in detail.

License status. We confirm that each therapist holds an active license by checking the license lookup tool operated by the issuing state board itself — not a copy of that data hosted somewhere else. We record the license number, the state, and the date we checked it.

Identity (NPI). Most licensed therapists in the United States have a National Provider Identifier, a unique number issued by the federal government and published in the NPPES registry. We match each profile’s NPI against that registry to confirm the profile describes the right person — the right name, credential, and location — and not someone else with a similar name.

Disciplinary and exclusion records. We check state board disciplinary records along with two federal databases: the Office of Inspector General’s List of Excluded Individuals and Entities (OIG LEIE) and the SAM.gov exclusion list. When we find nothing, we say so plainly — “no public disciplinary actions found as of [date]” — because a documented absence, with a date attached, is more useful to you than silence.

Credentials and specialties. Degrees, certifications, and areas of focus are recorded with the source they came from, typically the therapist’s own site or professional listings, and cross-checked where records allow.

Reviews and reputation. Where we summarize a therapist’s reputation, the underlying reviews are drawn from named platforms and cited, and presented in balance — strengths and concerns alike. We never write, invent, or paraphrase reviews into existence.

Every automated system we use operates under one hard constraint: it is not allowed to invent anything. A fact that can’t be backed by a source stays blank. Every material claim in a profile is stored with the link it came from and the date it was verified.

This is why some ThriveTalk profiles have gaps. A missing field doesn’t mean we didn’t look — it means we looked and couldn’t confirm it to our standard. We consider an honest blank far more trustworthy than a confident guess, especially on a page about a real clinician you may be about to trust with your mental health.

The most common failure mode in automated research is a system grading its own homework. We designed around it: the verification pass that audits a profile never sees the research notes that produced it. It receives only the finished draft, then rebuilds the case from scratch against primary sources.

That independence matters in three ways. Errors the fact-checker can prove with a primary source — say, a license number transcribed incorrectly — are corrected before the profile ever reaches an editor, and the correction is logged with its evidence. Anything the fact-checker is less than fully confident about (its confidence score falls below our threshold, or it flags a concern it couldn’t resolve, such as a state board website that blocks automated lookups) is escalated to a human editor with the specific concern spelled out, so a person performs the check the machine couldn’t. And if the fact-checker concludes a draft describes the wrong person entirely, or contains a claim that looks fabricated, the profile is rejected outright. It never reaches the site.

AI does the heavy lifting of gathering and cross-checking data at a scale no human team could match. But AI does not decide what you see. A human editor reviews each profile individually — with the full fact-checking report, every correction, every flagged concern, and every evidence link in front of them — and makes a per-profile decision to publish or reject. There is no bulk-approve button in our workflow, and there is no path by which a profile reaches the site without that human decision.

When a profile is published, we keep a permanent internal record of what was published, when, what it said, and which sources supported each claim. That audit trail is what lets us stand behind every page — and correct it quickly if something changes.

Verification isn’t a one-time event. Licenses lapse, therapists move, disciplinary records get updated.

Every material claim on a ThriveTalk profile carries a verification date, and each profile shows when it was last fully verified. On a recurring basis, we re-run the entire research and fact-checking cycle on published profiles. A refreshed profile goes through the same independent audit and the same human review as a new one — and if the fresh research can’t pass fact-checking, the existing page is not altered until a human resolves the discrepancy.

If you’re a clinician with a profile on ThriveTalk and something is wrong or out of date, we want to fix it — quickly.

Contact us with the correction and, where possible, a pointer to the authoritative record (your board listing, your NPI record, your current practice site). On any credible dispute, we re-verify the contested information against primary sources and update or annotate the profile promptly. Information you supply directly is noted as provider-supplied, and anything we look up on your behalf goes through the same verification standard as everything else on the site.

The same applies to removal requests: reach out, and a human will review and respond.

We’re often asked what’s under the hood. The honest answer is a mixture: AI research and fact-checking models, automated data collection across licensing boards and registries, proprietary tools we’ve built for cross-referencing sources and scoring confidence, and human editors who make every publication decision. Each layer exists to catch what the others might miss. The technology is what makes it possible to verify at this depth across thousands of profiles; the human review is what makes it trustworthy.