Help Center → Capture
Photo Quality and the Photo Score
There is no quality gate that blocks a photo. There is a score the AI gives each finding for the photo behind it — here's what it measures and how to read it.
5 min read
Photo Quality and the Photo Score
Start with what the system does not do: it does not screen your photos before analysis, it does not block a photo for being blurry, and there is no flag queue to clear.
Every photo you send to the AI gets analyzed. What comes back tells you how good the photo was.
The photo score
When the AI produces a damage finding, it also rates the photo that finding came from — a 0–100 photo score, built from three parts:
| Component | Weight | What it judges |
|---|---|---|
| Clarity | 0–40 | Sharpness and focus |
| Lighting | 0–30 | Exposure — neither blown out nor lost in shadow |
| Damage visibility | 0–30 | Whether the damage itself is actually legible in the frame |
Alongside the number, the AI writes a one-line reason — "Good clarity, slight angle" — that appears when you hover the score.
Where you see it
On the finding card in Analyze, and again in the photo lightbox, as Score 78. The color is the fast read:
| Score | Color | Read it as |
|---|---|---|
| 80–100 | Green | The photo carries the finding on its own |
| 60–79 | Blue | Solid. Fine for the file |
| 40–59 | Amber | Usable, but thin. Consider a supporting shot |
| 0–39 | Red | The finding rests on a photo that will not survive a reviewer |
The score belongs to the finding, not to the photo. The same photo can produce a strong finding and a weak one — the third component is about this damage in this frame.
Why it matters
The score is not decoration and it is not a grade on your camera work. It's a triage signal.
A high-severity finding sitting on a low-scoring photo is the weakest link in your file. It's the line an adjuster pushes back on, and it's the one you can still fix cheaply — while you're standing at the property.
That's the real workflow: run the analysis before you leave, sort by score, and re-shoot the low ones.
What drives the score down
Motion blur
The most common problem, and the most common cause is a ladder. Brace your elbow against your body before you tap the shutter, and tap to focus first on textured surfaces.
Backlighting
Shooting into the sun washes out exactly the surface detail the AI needs. Put the sun behind you, or wait for the cloud.
Underexposure on dark surfaces
Black shingles and dark siding fool automatic exposure. Tap the surface to set exposure on it.
Distance with no context
A close-up with nothing around it scores well on clarity and poorly on damage visibility, because the AI can't tell what it's looking at. Give every damage area an establishing shot, a mid-range, and a close-up.
Extreme angles
A surface shot from far off perpendicular is hard for the model to read and hard for a human reviewer to judge.
Photos of paper
Sketches, printed reports and handwritten notes are not field evidence. They belong in Notes or as an attachment, not in Capture.
Coverage is a different question
The photo score answers "is this photo good?". It says nothing about "did I photograph enough of the property?"
That's what the evidence completeness check is for, and it runs on every plan.
Frequently asked questions
Will a low-scoring photo still be analyzed? It already was — the score comes out of the analysis. Nothing is blocked or held back.
Can I override a score? There's nothing to override. It's an observation, not a gate. You decide what to do about it.
Can I mark a photo as high-priority for the AI? No. Run the analysis on the area and the AI reads what's there. To get better findings for an area, add better photos to that area.
My photos look fine on my phone but score low — why? An iPhone screen is bright and very high resolution, which flatters everything. The model reads the actual pixels. Check a doubtful shot on a monitor.
What score should I be aiming for? Don't chase a number. Chase this rule: no high-severity finding should be sitting on a red or amber photo when you close the job.
Capturing photos
Field technique and coverage, shot by shot.
AI damage detection
What the AI returns, and why you trigger it rather than it triggering itself.
Evidence coverage
Whether you documented enough of the property — the other half of the question.
Reviewing and managing media
Navigating, filtering and managing your media library.
Last updated: September 2026 · Questions? Email support@aiestimatics.com
Last updated: September 2026 · Feedback on this article