How wegotit.ai works, what it will and will not claim, and how to use the API.
1. Create an account. Sign up at wegotit.ai/signup using Google or email.
2. Research or verify. On the Search page choose Research to ask a question, or Verify a claim to fact-check a specific statement. Verify works best on specific factual claims, not opinions.
3. Review the evidence. wegotit.ai queries several live streams in parallel (web search, scholarly works, real-time news, public discussions, and video or images on media queries), then synthesises an answer where every factual sentence carries a citation marker. A text research query runs 6 streams on Free, 7 on Pro and 8 on Pro+.
4. Export and share. Export results as Markdown or PDF, or create a shareable link to your research.
Citations are built from the same retrieved set that is fed to the model, by a single builder. The model cannot cite a source that was not retrieved, so a fabricated citation is not something we detect and filter afterwards. It is something the pipeline cannot produce.
Above that, we fetch the full text of the top sources and locate the passage supporting each cited statement, then check each factual sentence in the answer against that passage. A sentence we cannot anchor to a quote is marked unsupported rather than presented as fact. Where coverage is too thin to answer, the tool declines instead of filling the gap.
Every answer carries a trust state, which is the honest description of how far that particular answer got:
| retrieved | Sources were found, but no answer was grounded in them. |
| cited | Every factual sentence carries a citation marker pointing at a real retrieved source. |
| evidence-linked | We fetched the source text and located supporting passages you can open and read. |
| claim-checked | Each factual sentence was checked against its passage and given a verdict. |
verified: true is set only at claim-checked, and it means every supported claim carries a quote that round-trips against the source text. It does not mean every claim was supported. When some claims are unsupported, the answer says so in the open rather than averaging it away.
There is deliberately no confidence score. A single number invites readers to trust a claim without opening the source, which is the behaviour this product exists to discourage.
The wegotit.ai API provides programmatic access to search and synthesis. All authenticated endpoints require a valid session token.
POST /api/searchSearch across multiple data sources.
Request body:
{
"query": "your search query",
"mode": "all" | "web" | "news" | "videos" | "images" | "social"
}POST /api/synthesizeGrounded synthesis with citations built from the retrieved set.
Request body:
{
"query": "your query",
"sources": ["web", "video", "news"],
"sessionId": "optional-session-id-for-follow-ups"
}Response:
{
"ok": true,
"data": {
"answer": "Synthesised answer where every factual sentence carries a [n] marker.",
"contractVersion": 1,
"trustState": "cited",
"verified": false,
"groundingCoverage": 0.92,
"citedIndices": [1, 2, 4],
"citations": [
{
"index": 1,
"title": "...",
"url": "...",
"source": "...",
"snippet": "...",
"sourceTier": "primary",
"sourceType": "news",
"purpose": "factual",
"textDepth": "full",
"fetchStatus": "ok"
}
],
"sources": { "queried": [...], "succeeded": [...], "failed": [...] },
"sessionId": "abc123"
}
}trustState and verified are described under How verification works.
groundingCoverage is the share of factual sentences carrying a valid citation marker. It is a citation-hygiene measure, not a truth score.
citedIndices lists which citations the answer actually used. Retrieved but uncited sources are returned too, and shown dimmed in the interface.
GET /api/healthPublic health check. Returns overall status, version, and the state of each downstream service.
Questions, bug reports, or something the tool got wrong? Email ask@wegotit.ai or use the contact form. We especially want to hear about answers that were confidently wrong.