How KnowThePerch handles sourcing, AI-generated content, fact-checking, and corrections. This is the reader-facing contract.
If you’re reading a species profile, state guide, or comparison page, the descriptive text was generated by a locally-hosted language model (Qwen3) based on verified data, then reviewed by a human editor before publication.
Descriptive paragraphs about behavior, habitat preferences, identification tips, and seasonal patterns.
Measurements, species counts, conservation status, frequency percentages, hotspot names, taxonomy, or any numerical claim. Those come from APIs.
Blog posts and birding guides are written by human authors without AI generation.
Measurements, taxonomy, conservation status, and observation data are validated against source APIs when each page is generated. AI-generated text is checked against the injected data for consistency.
If prose contradicts the source data — for example, claiming a species is “common” when eBird frequency shows it below 5% — the text is rejected and regenerated. The data always wins.
If you find incorrect information on any page — a wrong measurement, outdated conservation status, or a prose claim that contradicts the data — contact us with the page URL and the specific error.
We review correction requests within 5 business days. Confirmed errors are corrected at the source (database or prompt template), not patched in the rendered page. The page is then regenerated and the updated date is revised.
This means the same error cannot recur on future pages built from the same data source.
Report an error →