Take or upload a photo of a species you've seen, and HydroVision AI suggests what it is. It works on mammals, birds, fish, invertebrates, plants, and fungi. It's an aid, not a verdict — always verify the result against the species card before submitting.
From the Biodiversity Monitor:
- Tap the purple HydroVision AI™ card on the form
- Choose Take Photo (uses your device camera) or Upload (from your gallery)
- The photo uploads and the AI analyses it. This takes 5–15 seconds.
- A suggestion appears with the species name and confidence level
- Tap the suggestion to accept it — the species is pre-filled into the form
You can also use HydroVision AI as a standalone identifier without submitting a sighting. It's useful when you just want to know what you're looking at.
Good on clear photos of common species. Less reliable on:
- Poor-quality photos — blurry, dark, distant, obscured
- Juveniles that look very different to adults
- Species with many visually similar relatives (some warblers, some grasses, some fungi)
- Images with multiple species in frame — the AI will pick one
- Species not in its training data — rare, very regional, or recently-arrived invasives
The AI gives you a suggestion with a confidence level. Treat high confidence as "probably right, verify anyway." Treat low confidence as "interesting guess, look closely before trusting."
The app shows you a reference image and conservation status for the suggested species. Look at them. If what you photographed doesn't match what the reference shows, the AI is wrong and you should either:
- Search manually by name
- Try a different photo (better angle, better light)
- Record at a broader level (e.g. "Warbler sp." instead of guessing a specific warbler)
It's fine to say "I'm not sure." A record at family level is more useful than a confidently-wrong record at species level.
Because the dataset only works if it's trustworthy. Every photograph is read in the background and the reading is recorded alongside your own count. If the two disagree dramatically, that's flagged. The whole point is to have a record that regulators can trust — which means we can't let the AI be the final say.
Your observation, verified by photographic evidence, cross-checked by AI, located precisely by what3words — that's the chain of custody that makes angler and naturalist data count.
- The check is not done by us. The photo leaves your device, goes through our server, and is read by an outside AI service working under contract to us. Our privacy policy names that service and says what it does with the photo and for how long. If you would rather it did not happen, do not use the identification check: every form takes your report without it.
- The photo goes as it is. We do not take the camera data out of the file first, so if your camera wrote a GPS position into it, that position goes with the photo. A photo attached to a report is resized and saved again on your phone before it is sent, which does not carry that data across, but the identification check does not do that.
- We do not train anything on your photos. What the AI service itself may keep, and for how long, is set out in the privacy policy linked above, because that is their term and not ours to promise.
- A photo on a record is public. Photographs attached to reports are stored with the record and shown as evidence on the Live Data Map, with no name against them.