About Know That Bug

Who writes this

Jamie Rand. I build and maintain Know That Bug, and every page on it carries my name in its markup rather than a company byline or an invented persona.

I am not an entomologist, and this page is the wrong place to imply otherwise. What I do is narrower and, I think, more useful to state plainly: I check every identity and danger claim on this site against a named reference source before it publishes, I record which source and on what date at the foot of every page, and I publish what the identifier gets wrong alongside what it gets right.

That boundary is deliberate. A site telling you whether something can hurt you should be judged on whether its claims are checkable, not on a credential in a byline — so the method is on how we identify, the measured accuracy including the failures is on accuracy & limits, and the sources are on every record.

If something here is wrong, I would genuinely rather know. The correction route is at the bottom of this page.

Why this exists

Almost every bug identification site answers a question nobody asked. You find something in your bathroom at midnight and you want to know whether to worry — not its taxonomic order. So every result here leads with the danger answer, then the evidence behind it, then what else it could have been.

That last part matters more than it sounds. “This is a cellar spider” is only reassuring if you know we considered a brown recluse and ruled it out.

What we publish

85 species records across 12 groups, each carrying the reference sources its identity and danger claims were checked against, with the date they were checked. A record cannot be published without them — that is enforced when the site is built, not left to somebody remembering.

We also publish how accurate the identifier is, measured against 168 expert-verified photographs, including the cases it still gets wrong. Most tools in this space show a confidence percentage they cannot justify. We show three bands and the measured accuracy of each.

What we refuse to do

• No medical advice. We identify. If you have been bitten or stung and symptoms develop, see a doctor.
• No pest control advice. Identification, not extermination.
• No invented numbers. No fabricated species counts, no fake live sightings, no confidence percentages.
• No invented expertise. There is no fictional entomologist behind this site, and no borrowed credentials. There is a named person, sources on every claim, a stated method and published accuracy.

How a record gets written

Every species page follows the same order, and the order is the point: the danger verdict, then the features you can check it against, then where it lives and how it behaves, then what it is mistaken for. Taxonomy goes in the body, never the headline — somebody standing in a bathroom at midnight is not asking about family-level classification.

Before a record can be published it has to name the reference sources its identity and danger claims were checked against, with the date. That is enforced when the site is built rather than left to discipline: a record without sources fails the build. The same check requires every “commonly confused with” link to resolve and to be reciprocal, so the comparisons you are offered are real rather than decorative.

We write to answer questions, not to reach a length. If a section could be removed without losing an answer, it should be.

How the danger call is protected

The identifier is a vision model, and vision models fail in a specific way: they assign an animal to the wrong group before any reasoning about danger happens. A widow read as an orb weaver is never assessed as a widow.

So the danger verdict does not depend on the model alone. A fixed rule in our code reads the answer and refuses to publish “harmless” when the answer’s own shape cannot support it — a low-confidence identification in a group that recluses, widows or bark scorpions are mistaken for gets an explicit “we can’t rule this out” instead. It never invents danger; it only withdraws a claim of safety, and it says which claim it withdrew and why.

The measured effect, and the numbers behind it, are on accuracy & limits.

Who writes this

Know That Bug is published by the organisation of the same name. There is no named expert on these pages, and that is deliberate: attaching a fabricated persona to safety claims is a common practice in this niche and it is not one we are going to copy. What we offer instead is checkable — the sources on every record, the method above, and accuracy figures measured against expert-verified photographs and published in full, including the failures.

If a named editor stands behind the site in future, they will appear here and in the page markup at the same time, and not before.

Corrections

If something here is wrong — particularly a danger claim — we want to know, and we will fix it and update the checked date on the record. Accuracy on this site is a safety matter, not a matter of pride.

Read next: how we identify · accuracy & limits · privacy