How to Verify an AI Price Estimate Before You Buy or Sell

Step 1: Confirm the Exact Identity

Match brand, model, year, edition, size, material, serial or reference number, card number, storage capacity, colourway, and included components. Compare the proposed identity against the labels and the physical object in front of you rather than accepting it because it sounds right. If the app has identified only a broad category, do not then use a narrow price range derived from it — the precision of the range must never exceed the precision of the identification. A plausible wrong identity is the most dangerous error in the whole process, because every subsequent comparison will look orderly and consistent while answering a question about a different object.

• Brand, model, year, and edition.
• Size, material, and colourway.
• Serial, reference, or card number where one exists.
• Configuration such as storage capacity.
• Included components and accessories.
• Range precision must never exceed identification precision.

Step 2: Define the Real Condition

Record whether the item works, and whether it is complete, sealed, graded, altered, repaired, restored, locked, authenticated, or missing accessories. Photograph the defects, and distinguish clearly between what you tested and what you assumed. Use a category-specific condition standard where one exists, because those standards are the vocabulary your comparables are described in and using your own words makes matching harder. Do not compare your damaged, incomplete, raw, or repaired item against pristine, complete, graded, or professionally restored examples simply because the listing titles are similar — title similarity is the trap, and condition is where most of the price difference lives.

• Functional status: tested, untested, or known faulty.
• Completeness, and exactly which accessories are absent.
• Sealed, graded, raw, altered, repaired, or restored.
• Account or activation lock status for devices.
• Use the category's own condition vocabulary.

Step 3: Inspect the Evidence Type

Determine what each source actually is: a completed sale, an active listing, a retail offer, an auction result, a price guide, or a loosely related page. A completed transaction is usually stronger resale evidence than an asking price, but only when the match is close — a distant completed sale can be weaker evidence than a near-identical active listing, so type alone does not settle it. Check the date, currency, region, shipping arrangement, buyer's premium, lot size, and whether the result is still accessible for you to re-examine. Do not blend different evidence types together without labelling them, because a blended average of a retail price and a completed sale describes nothing that exists.

Step 4: Collect Five to Ten Close Sales

Search the exact identifiers and retain five to ten recent completed sales where the market supports it. Three reliable matches can serve as a minimum reference, but more is better whenever condition varies across the examples you find. Keep a simple table recording date, platform, item details, condition, sale price, shipping, and any notes — the act of writing them down is what stops you remembering the set as more favourable than it was. If the market is thin, widen the time window cautiously and lower your confidence to match, but do not substitute distant lookalikes merely to reach a target count. A comp set padded to look substantial is worse than an honestly small one.

• Five to ten recent completed sales; three as a hard minimum.
• Record date, platform, details, condition, price, and shipping.
• Widen the time window before loosening the match specification.
• Never pad the set with distant lookalikes to reach a count.
• Write it down — memory flatters the evidence.

Step 5: Remove Mismatches With Reasons

Exclude a result only for a documented reason: wrong edition, different size, missing or extra components, counterfeit or reproduction, bundle, parts-only status, different grade, unrelated region, or corrupted data. Keep ordinary high and low outcomes, because those represent genuine market spread and removing them manufactures a confidence the evidence does not support. The standard to hold yourself to is that a clean verification record should let another person understand why every retained comparable belongs in the set and why every excluded result does not. If you cannot articulate the reason for an exclusion, that is a strong signal you are excluding it because it is inconvenient.

• Valid reasons: wrong edition, size, grade, region, or components.
• Valid reasons: counterfeit, reproduction, bundle, parts-only, corrupted data.
• Not a valid reason: the number is inconvenient.
• Keep ordinary high and low results — that is real spread.
• If you cannot state the reason, do not exclude it.

Step 6: Calculate the Median and Range

Sort the retained total prices and take the middle value as your reference centre. Report the minimum, maximum, median, sample count, and date window rather than a single falsely precise number, since the spread and the sample size carry as much decision-relevant information as the centre does. Then compare the AI range against this evidence and ask three specific questions: does it contain the median, does it overlap the normal spread, or does it sit entirely outside the observed range? A broad estimate range may be perfectly honest while being less actionable; a narrow one requires stronger and more consistent matches to justify, and should be treated sceptically when the underlying evidence was thin.

Step 7: Adjust for the Actual Decision

For resale profit, subtract platform fees, payment costs, shipping, packaging, repairs, cleaning, authentication, returns, taxes where applicable, and the acquisition cost itself. For a purchase decision, include the downside case if the item is untested or slow-moving, rather than modelling only the outcome you want. For local goods, account for transport and storage, both of which are real even when no invoice arrives. And for insurance, estate, legal, tax, donation, rare, luxury, or authenticity-sensitive decisions, stop treating a consumer estimate as the final authority and consult a qualified professional — the verification process in this guide is designed to tell you when you have reached that point, not to substitute for it.

Read next:Resale profit calculator ·How PriceSnap estimates resale value