Honest answers. Even when the honest answer is "no."

No — and any service telling you otherwise is lying. Final grades are decided solely by the human graders at PSA, BGS, SGC, CGC, or TAG. We don't grade your card; they do. What we provide is independent, math-backed decision support: probability distributions across each grade tier (e.g., "PSA 10: 41% · PSA 9: 38% · PSA 8: 18%"), confidence scores calibrated against PSA Pop Reports and millions of historical sales, sub-grade analysis (corners, edges, centering, surface), market comps, and the EV math that tells you whether submitting is worth the fee. Human graders disagree on the same card all the time — that's why crossovers and regrades exist. Our value is helping you make a smarter pre-submission decision, not promising an outcome only the third party can deliver. If a card comes back lower than predicted, we want to know — every miss feeds the model.

We target 70%+ accuracy within one grade tier on top-population cards across each category. Confidence scores tell you how much to trust each prediction — a 92% confidence prediction has a different track record than a 64% one. Low-confidence predictions are flagged as "submit at your discretion." We publish our calibration metrics quarterly.

Public PSA Pop Reports, every verified-grade auction sale we can ingest (eBay, Goldin, PWCC, Heritage, Fanatics Collect, Whatnot), every documented grading-service scandal and case study, expert grading literature, sub-grade decompositions from Beckett, and our own ongoing user-prediction-vs-actual feedback loop. 5,400+ training sources, 600+ primary documents.

Sometimes. The AI is best at centering, surface flaws visible in photos, factory miscuts, and counterfeits with known fingerprints. It is worse at: very fine print quality issues that require a 10× loupe, ink errors invisible at normal scan resolution, and condition discrepancies that depend on lighting angle. We recommend you photograph in daylight at 45° with the card sleeved-but-not-cased.

Strong on the highest-faked cards in each category — Base Set Charizard, LOB Blue-Eyes, 1986 Fleer Jordan, Bowman Chrome high-end rookies, OP08 Manga Rares, Beta/Unlimited Power Nine. Weaker on freshly-discovered fake operations we haven't seen yet (you should still consult a trusted authenticator for purchases over $5K).

Six options: Free trial (3 AI analyses, one-time, no card on file), Credit packs ($29 / $99 / $299 for 10 / 50 / 200 credits, never expire), Pro ($49/mo unlimited), Pro Annual ($490/year — 2 months free), Lifetime Founder ($1,499 once, capped at 500 buyers), and Shop tier ($199/mo for 5 seats). Full breakdown on the pricing page.

No catch — but there's a hard cap of 500 buyers. After that, the tier closes permanently at $1,499. We may open a second cap at $1,999 later, but the original 500 founders lock pricing on every future product (mobile app, data licensing, courses, etc.) — that's unique to this cohort.

Credits: 14-day refund on any unused portion. Annual: 14-day refund less consumed months at monthly rate. Lifetime Founder: 30-day refund, full $1,499 back, no questions. Shop: cancel anytime, no clawback. Single Pro month: cancel anytime, charges stop.

Yes — 20% off Pro monthly with .edu email or military ID verification through ID.me. Email [email protected] after signing up and we'll apply it manually.

Most collectors don't stay in one category forever. The Pokémon collector also holds a Charizard, also has some Lorcana, also bid on a Beta Underground Sea once. Splitting AI tools across 6 brands meant you'd pay six subscriptions and learn six interfaces. CardPrepAI is one engine, one subscription, every category — same data quality, same UI, every card you'll ever own.

Pokémon TCG (WOTC + TPCi + Japan), MTG, Yu-Gi-Oh!, One Piece, Lorcana, Star Wars Unlimited, Riftbound, Sorcery, Digimon, Flesh and Blood, Dragon Ball Super, all major sports (NFL / NBA / MLB / NHL / soccer / boxing / MMA / golf / tennis / motorsport), wrestling, racing, non-sport (Garbage Pail Kids, Marvel, etc.), and sealed product across every category. The cascade selectors on every tool guide you through narrowing to the right one.

Pre-war (1900–1948) — T206, Goudey, Play Ball, early baseball/boxing — is fully supported. Vintage baseball, basketball, hockey, football from the 1950s–1980s is fully supported. The cascade includes a "Pre-war (1900–1948)" year bucket.

Japanese Pokémon, OCG Yu-Gi-Oh, German MTG, Spanish soccer, Italian Panini — all supported. The Sentry and Predict tools both have a Language dropdown.

You'll land on your member dashboard. Hit "Predict a new card →," upload front and back photos, get a full prediction in 25–45 seconds — PSA / BGS / SGC / CGC / TAG grades, sub-grade analysis, market comps, submission-tier recommendation, net EV math, alteration screening, and a personalized next-action checklist. Every prediction stays in your dashboard history.

Yes. Export to CSV from your dashboard anytime. We never sell individual user data. We do use aggregated, anonymized prediction data to improve the model and (eventually) license to hobby publications.

Photos are stored on Cloudflare R2 (encrypted at rest), used only for your prediction history, and retained for 24 months by default. You can delete any prediction at any time from your dashboard. We never publish your photos publicly or share them with grading services.

No. CardPrepAI is digital-only. We don't take possession of cards, we don't ship cards to graders, we don't have a vault. Concierge tools generate paperwork and recommendations; you ship to PSA / BGS / SGC / CGC / TAG yourself. This is by design — your cards stay in your custody.

$19/card. A real human (former TPG grader or experienced high-volume submitter) reviews your photos plus the AI's prediction, then returns a written verdict within 24 hours: submit, re-shoot the photo, skip, or change tier. Useful for high-value cards where you want a second pair of eyes before paying for grading.

$99 for 30 minutes (or $199 for 60). Live video call with a grading strategist. Screen-share your batch, we run net-EV across all 5 graders, optimize tier mix, flag risks, talk timing. Best for collectors prepping a batch of 20+ cards or anyone making their first major submission.

30% recurring commission for every month a referred collector keeps their CardPrepAI subscription active. 60-day cookie. Monthly payouts on the 15th. If the referred customer cancels, the commission stops for that month and resumes if they come back. Full details + apply →

Where we link out to Amazon (submission kit) or marketplaces (eBay, Goldin, PWCC, etc. in the Sniper/Listing tools), we attach our affiliate ID to the URL. You pay the same price as anywhere else — we earn a small commission from the marketplace. This is disclosed with FTC-compliant "sponsored" tags. Helps us keep the free tier free.

Yes. eBay, Goldin Marketplace, PWCC Vault, Heritage Auctions, Fanatics Collect, Whatnot — across every category we support. Set a watchlist filter (e.g., "PSA 10 Charizard alt arts under $1,200") and the AI watches in the background. Pro tier includes sub-second snipe bidding.

Break Co-pilot covers booster boxes, ETBs, UPCs, Premium Collections, hobby cases, hanger boxes. AI tells you whether to open, hold, or break-buy based on PSA-graded pull math plus sealed-product appreciation curves. Cross-TCG and sports.

Daily PSA Pop Report monitoring on every card you've added to your portfolio. Alert thresholds: net-new-graded count exceeds your weekly threshold, gem-rate shift, registry-significant changes, suspected pop-doctor activity.

Email [email protected] — typical response under 12 hours, Pro and above get 4-hour SLA. Founders and Shop tier get a dedicated Slack channel.

Submit your actual graded result through the dashboard's "Report actual grade" button on any prediction. This trains the model and earns you a $5 credit on Pro subscriptions (capped at $50/month). We take misses very seriously.