AI card grading, more accurately called AI pre-grading, estimates a trading card's condition from photos before you pay a grading company. Software finds the card in the picture, measures its centering, inspects the corners, edges and surface for defects, and turns what it finds into a score on the familiar 10-point scale. It is a filter, not a certificate: the result tells you which cards are worth sending, but the only official grade is the one a grading company gives after handling the card.
What AI pre-grading is (and what it is not)
Professional graders such as PSA, Beckett (BGS) and CGC examine each card by hand, under magnification, and seal it in a slab with a grade that the market recognises. That service costs a fee per card and usually takes weeks. AI pre-grading does a first pass in minutes, at home, so you only pay for the cards with a realistic chance of the grade you want.
- It is an estimate of condition, with sub-scores that show where a card loses points.
- It is not an official grade, an authentication, or a guarantee of what a grading company will decide. Different companies also apply their own standards.
How AI card grading works, step by step
Every tool differs in detail. This is exactly how TCGAI.PRO does it (the full version, with the rules, is in our methodology):
- Photo check. The software rejects photos with no card, heavy blur or unusable exposure, and flags softer problems such as glare as warnings.
- Find and straighten the card. The card's outline is detected and the image is corrected to the standard 63 × 88 mm shape, so an angled photo is measured as if it were scanned flat.
- Measure centering. Border widths are measured in pixels on both faces and converted to a ratio such as 60/40, then to a centering sub-grade.
- Inspect corners, edges and surface. Zoomed crops of the corners and edges go to a vision model, which lists each defect with its type, position, severity and size.
- Apply fixed rules. Published rules turn defects into sub-grades. The overall score is the lowest sub-grade, because one weak area caps a card.
- Report the uncertainty. A confidence range (for example 9 ± 0.5) widens when the photos are poor or the back is missing.
AI pre-grading versus professional grading
| AI pre-grading | Professional grading | |
|---|---|---|
| What you get | An estimate with sub-scores | An official grade in a sealed slab |
| Time | Minutes | Usually weeks, depending on the service level |
| Cost | Free tier, then about $0.10 to $0.23 per scan on paid plans | A fee per card, plus shipping and insurance |
| Handling | The card never leaves your hands | The card is shipped and examined in person |
| Recognised by buyers | No: it is a screening tool | Yes |
| Main limit | Can only judge what the photo shows | Cost and waiting time |
For current fees and turnaround times, check each company's own pages: PSA services, the Beckett grading scale and the CGC Cards grading scale. Our centering tables follow PSA's published centering tolerances.
What AI can and can't detect
- Reliable: centering on cards with a visible border, corner whitening and rounding, edge chipping and whitening, and obvious surface damage such as scratches, dents, creases and stains.
- Harder: fine print lines, light holo scratches and surface gloss loss. They often show only under angled light, so a single flat photo can miss them.
- Out of reach: anything that needs the card in hand, such as thickness, trimming checks or confirming that a card is genuine. For fakes, see our guide to identifying fake cards.
When is a card worth submitting?
Grading pays when the grade you can realistically expect is worth more than the card raw plus the grading cost. Pre-grading gives you that realistic expectation. Some context from the 7,270 cards graded on TCGAI.PRO between December 2025 and October 2026:
- Only 20.6% scored a perfect 10, while 60% scored 9 or higher.
- Pokémon cards from 1999 to 2002 averaged 6.42, against 8.79 for cards from 2023 onwards.
- Among imperfect cards, edges and surface were most often the weakest sub-grades.
A practical rule: submit when the score is high, the confidence range is narrow, and the price gap between the expected grade and the raw card clearly covers the fee. Rescan with better photos when the range is wide. Keep raw anything that is likely to land a grade below its raw value.
How to get an accurate result
- Take the card out of its sleeve or toploader: plastic adds glare and hides edges.
- Use even, diffuse light and a plain dark background.
- Shoot straight on, fill most of the frame, and keep the card in focus.
- Always add the back: without it, the result is less certain.
Our photography guide covers lighting and angles in detail, and the condition guide explains what each sub-grade looks for.
Frequently Asked Questions
Is AI card grading the same as PSA or BGS grading?
No. AI pre-grading gives an estimate from photos; PSA, BGS and CGC give official grades after examining the card in person. The TCGAI Score is not affiliated with or endorsed by any grading company.
How accurate is AI card grading?
It is as accurate as the photos allow. Measured centering is precise; surface defects that only show under angled light are the most common misses. That is why every TCGAI.PRO result shows a confidence range. Our article on AI grading accuracy goes into detail.
What does the TCGAI Score measure?
Four sub-grades on each face: centering, corners, edges and surface. The overall score is the lowest of them, since one weak area usually caps a card's grade.
Can I pre-grade a card that is already in a slab?
Yes, although the slab adds reflections. Collectors do it to judge whether a crack-and-resubmit could earn a higher grade.
Is AI pre-grading free?
TCGAI.PRO includes two free scans a month on every account, enough to try it on your own cards before choosing a plan.
Pre-grade a card now and see which sub-grade decides its score.