How to Grade a Pokemon Card with AI

The grading pipeline begins the moment you open the CardGrading.app camera. Our guided capture system helps you position your card in the frame, checks for adequate lighting, and captures a high-resolution photograph at up to 4K quality.

Why does resolution matter? Professional graders examine cards under magnification. A 4K image gives our AI enough pixel density to detect micro-level imperfections - hairline scratches, tiny dents, and faint print lines - that a standard phone photo would miss. The app also corrects for angle distortion so every card is analyzed from a perfectly flat perspective.

Unlike a quick visual check, our AI grades both sides of the card independently when you include a back photo, the way PSA evaluates a card. The front is evaluated for surface scratches, print defects, ink inconsistencies, and holo pattern damage. The back is analyzed for whitening along edges, indentations, and color fading.

Many collectors overlook the back of a card when self-grading, but PSA assigns roughly equal weight to both sides. It applies PSA's published grading criteria - the same centering, corner, edge, and surface checks a grader uses, now running as AI vision. When you include a back photo, this dual-side approach helps you catch what's visible in your photos - especially the subtle edge whitening on card backs that often drops a grade from PSA 10 to PSA 9.

Our centering overlay calculates the centering ratio on the front of the card - and on the back when you include a back photo. It measures the border width on all four sides and computes left-right and top-bottom centering percentages - the same 60/40 and 55/45 thresholds PSA uses.

Centering is one of the most objective grading criteria and one of the easiest for AI to measure. A card with 50/50 centering on both axes is perfect. PSA allows up to 60/40 for a 10 and 65/35 for a 9. Our overlay gives you the percentages so you know where your card stands before you submit. Read our full centering guide for more detail.

Each of the four corners and each of the four edges is cropped, isolated, and magnified with zoomed-in analysis. The AI inspects every corner individually and every edge individually for whitening, dings, soft rounding, chips, rough cuts, and dents - the most common reasons cards lose grades.

Corners and edges are the most vulnerable parts of any trading card and often the deciding factor between a PSA 9 and a PSA 10. Even a tiny speck of whitening on a single corner can drop your grade. By zooming into each corner and each edge separately, our AI can surface defects visible in the submitted photos - checking the same corners and edges PSA weighs most heavily.

The AI combines data from all previous steps - surface quality on both sides (when a back photo is included), centering ratios, individual corner conditions, and individual edge wear - and applies PSA's published grading criteria to produce a predicted PSA grade from 1 to 10.

You receive an overall predicted grade plus individual sub-grades for centering, corners, edges, and surface - mirroring the four categories PSA evaluates. This breakdown tells you exactly which area is strongest and which might be holding your card back. It applies the same corner, edge, surface, and centering checks that PSA's published criteria describe. Learn more about what each grade means in our grading scale guide.

Once your card is identified, it's matched against multiple trading card pricing databases to pull live market prices. You see four values: what your card is worth raw (ungraded), at PSA 9, at PSA 10, and an interpolated value at your predicted grade.

Prices are sourced from PokeTrace (which aggregates data from TCGPlayer, eBay, and CardMarket including actual PSA sold prices), eBay sold listings, and TCGPlayer. Coverage is strongest for Pokemon, where we can usually show values raw, at PSA 9, and at PSA 10. Sports cards and other TCGs are graded exactly the same way, but a value estimate may not be available for every card. All prices are estimates based on aggregated recent sales data. Actual sale prices vary based on listing format, condition, and market demand.