Home · Business · Operations & Logistics · Quality control

PARTLY

As of 13 August 2026, AI can only partly check product dimensions from images.

This still needs a person who signs their name to it.

Can you do it?

5 minutesto a draft.

30 minutesto something you’d act on.

Cost, all in£0

Skill neededchat-fluent

Who has to check ita colleague

What the alternative costsNo priced human or specialist measurement alternative is provided in the available tool data.

If this goes wrong: a perspective error or missing scale makes an acceptable product look compliant, or a compliant product look defective, and the mistake enters your quality-control decision.

What to actually do

  1. Hand it to a person

    The route this page recommends

    A person who owns the outcome does this end to end, worth it when the failure is dear.

  2. Use a tool built for this

    Second choice

  3. Do it yourself

    The distant third

    A chat interface, chat-fluent skill, and roughly 30 minutes until you can act on the result.

    How to actually do it

    1. Open the product specification or quality standard and list each required dimension, unit and permitted tolerance.
    2. Photograph the product with its edges visible, the camera square to the measurement plane, and a ruler or known-size calibration marker in the same plane.
    3. Take additional front, side and top views where one image cannot show the required dimension without obstruction or perspective distortion.
    4. Paste the prompt into ChatGPT, Claude or Gemini and attach the images, then replace the bracketed fields with the dimensions and tolerances from your specification.
    5. Save the model's measurement table with the image or view cited for each result, and mark every low-confidence or indeterminate result for physical checking.
    6. Measure flagged dimensions on the product with the approved ruler or callipers, compare those readings with the specification and tolerance, and record the checked result in your quality-control system.
    7. Ask a responsible colleague to confirm the image evidence and physical readings before releasing, rejecting or reworking the product.

    Prompt

    Check the attached product images for these dimensions: [list the required dimensions and acceptable tolerances]. The images include [describe the ruler, calibration marker or known-size reference]. Use only visible evidence. Do not infer absolute dimensions if the scale, camera angle or object edges are unclear. For each dimension, report the estimated value and unit, the tolerance comparison, the image or view used, the reference used for scale, and a confidence level of high, medium or low. Identify perspective, occlusion, lens distortion and other reasons the estimate may be wrong. If the images do not support a reliable measurement, say 'cannot determine from these images' and state exactly what additional photograph or physical measurement is needed. Do not approve or reject the product solely from an uncertain estimate.

    Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.

What it gets wrong

  • AI cannot establish absolute scale from an ordinary image that contains no known-size reference.
  • AI cannot correct reliably for hidden edges, lens distortion or a camera that is not square to the product.
  • AI cannot replace a calibrated ruler, callipers or a controlled measurement fixture for an acceptance decision.
  • AI cannot take the physical measurement or decide whether an ambiguous defect meets your internal quality standard.
  • AI cannot carry responsibility for releasing or rejecting stock.

What caps this at PARTLY: physical presence, verification cost and stakes of error.

How we scored this

Five axes, each scored nought to two by hand: ten means AI carries the task cleanly, and the thresholds that turn a total into YES, PARTLY or NO are published in the methodology. Each axis name links to its definition.

AxisScore (0–2)
Output1
Inputs1
Verification2
Liability1
Effort delta1
Total6 / 10

The methodology and its thresholds are published in full.

FAQ

Can ChatGPT measure an object from a photo?
It can estimate dimensions if the photo includes a visible ruler, calibration marker or other known-size reference and the relevant edges are clear. Without a scale reference, it cannot reliably provide an absolute measurement.
How accurate is AI at measuring product dimensions from images?
Accuracy depends on scale, camera angle, image quality, lens distortion and whether the full edges are visible. Treat the result as a screening estimate, then check dimensions that affect acceptance with approved physical measuring equipment.
Can AI check whether a product meets its size tolerance?
Yes, it can compare an image-based estimate with the tolerance you provide. It should not make the final pass or fail decision when the estimate is uncertain, the image lacks calibration or the dimension is safety or contract critical.
What do I need to photograph product dimensions accurately?
Use clear images with the product edges visible, the camera square to the relevant surface and a ruler or known-size marker in the same plane. Take separate views for obstructed dimensions and keep the original images with the inspection record.

Nearby answers

All Quality control rulings

Assessed by gpt-5.6-luna (gpt-5.6-luna) on 2026-08-13, second-checked by an independent model. Wrong somewhere? Email [email protected] and it gets re-checked.

The newsletter

AI news, new answers and product picks, straight to your inbox.