Patent checkup

Enter a patent number or upload the patent PDF. Fixed rules turn public data into a checkup score, and each part of the score is explained.

The rules are published. For design-around difficulty, two AI models from different vendors each read claim 1 and check, feature by feature, whether another way could serve the same purpose; a feature they disagree on counts as half. When you check by patent number the judgments are archived, so the same archived judgments give the same score under the same rules version. The score reflects structural features found in the public document. It is not a conclusion about the patent’s legal effect.

Use the patent number from a Chinese certificate (starting ZL), or a publication number from the US, Europe, China and more. Spaces and commas are fine.

How the score works

Three dimensions, 100 points in total, the bulk of them for design-around difficulty. Design-around difficulty is judged feature by feature by two AI models from different vendors, and without it there is no total. The other two dimensions are computed entirely by fixed rules; if the information one of them needs cannot be read, it is not scored and the total is scaled to 100 over the dimensions that were scored. When checked by number, the same archived judgments give the same score under the same rules version. Rules version 6.

Data comes from the public patent document. A lookup by number reads the public patent page, an upload uses only the patent number or claims found in the file, and pasted claims use only the text you paste.

  1. Scope of protection

    5 points

    How broadly claim 1 is drafted: fewer and shorter limitations mean a wider scope.

    Number of limitations in claim 1

    • Under 4 limitations: 3 points
    • 4 to under 6 limitations: 3 points
    • 6 to under 8 limitations: 2 points
    • 8 to under 10 limitations: 1 point
    • 10 to under 13 limitations: 1 point
    • 13 limitations or more: 0 points

    Length of claim 1 (Chinese by characters and English by characters, with different bands)

    Chinese claims

    • Under 151 characters: 2 points
    • 151 to under 351 characters: 1 point
    • 351 to under 501 characters: 1 point
    • 501 characters or more: 0 points

    English claims

    • Under 501 characters: 2 points
    • 501 to under 1101 characters: 1 point
    • 1101 to under 1501 characters: 1 point
    • 1501 characters or more: 0 points
  2. Fallback positions

    5 points

    How many positions are left to fall back on if claim 1 is challenged: the number of dependent claims, and whether the independent claims cover more than one category such as product and method.

    Number of dependent claims

    • Under 3 claims: 0 points
    • 3 to under 10 claims: 1 point
    • 10 to under 15 claims: 2 points
    • 15 to under 20 claims: 2 points
    • 20 claims or more: 3 points

    Categories covered by the independent claims (product, method, use)

    • Under 2 categories: 0 points
    • 2 categories or more: 2 points

    This needs the complete, numbered claims to count dependent claims. If only claim 1 is pasted, this part is not scored.

  3. Design-around difficulty

    90 points

    Starting from the features of claim 1, whether another way could serve the same purpose for each feature, or whether it could be left out. The higher the weighted count of such features, the lower the score. The percentage is shown for context only.

    Weighted count of claim 1 features flagged as replaceable or removable (a feature only one model flags counts as half)

    • 0 weighted features: 90 points
    • 0.5 weighted features: 80 points
    • 1 weighted features: 70 points
    • 1.5 weighted features: 60 points
    • 2 to under 3 weighted features: 50 points
    • 3 to under 4 weighted features: 40 points
    • 4 to under 5 weighted features: 30 points
    • 5 to under 6 weighted features: 20 points
    • 6 to under 7 weighted features: 10 points
    • 7 weighted features or more: 0 points

    Each AI model gives each feature one of three judgments

    • Hard to avoid: no other way serves the same purpose, and the feature cannot be left out.
    • Removable: the product still serves the same purpose without this feature. Counts as a bypassable feature.
    • Replaceable: another way, outside the literal wording of the claim, serves the same purpose. Counts as a bypassable feature.

    Whether each feature could be replaced by another way that serves the same purpose, or left out, is judged one by one by two AI models from different vendors, each reading claim 1 on its own without seeing the other. A feature both models flag counts as one, and a feature only one model flags counts as half. The published bands turn this weighted count into points. The percentage is shown for context only. The bands help identify points to review. They do not give a probability of successfully designing around a claim. How claim 1 is split into features can affect the score. If this AI step does not complete, there is no score this time. If only one model answers, each feature it flags counts as one, the page says this time was not cross-checked, and a later checkup fills in the other model. When you check by patent number each model’s judgments are archived. The same archived judgments give the same score under the same rules version.

Grades

  • 72 and above:Solid
  • 58 to 71:Good
  • 45 to 57:Fair
  • Below 45:Needs work

Frequently asked questions

How is the score worked out?
The claims are read from the patent’s public document, and fixed rules published on this page turn them into points. The bulk of the points are for design-around difficulty: two AI models from different vendors each read claim 1 and check, feature by feature, whether another way could serve the same purpose, a feature they disagree on counts as half, and the published bands then turn the weighted count into points. A higher count means fewer points. The percentage is shown for context only. The bands help identify points to review. They do not give a probability of successfully designing around a claim. Scope and fallback positions come from the features, length and dependent claims. A part that could not be read is not scored, and the total is scaled over the rest. When checked by number, the same archived judgments give the same score under the same rules version.
Does a high score mean the patent is safe to rely on?
No. The score reflects structural features that can be read from the public document, plus two AI models’ judgments of claim 1 feature by feature. It is not a conclusion about the patent’s legal effect or whether its scope would be upheld. That depends on the product and the prior art, and a lawyer judges it.
Do I need to pay or sign up?
No. The checkup is free and needs no account. Your details reach the firm only if you choose to have a lawyer look at the patent and send the form.
Are the files I upload or the claims I paste stored?
No. An uploaded file is used only to read it once and is then discarded. Files and pasted claims are not stored, are not written to logs and are not sent with the contact form. To judge design-around difficulty, claim 1 is read by two AI models from different vendors. The China site uses two models inside China and sends nothing abroad.
Why does the China site ask me to upload the patent document?
Looking a patent up by number needs a public patent database outside China, so the China site does not offer it, and a Chinese patent certificate does not contain the claims. Upload the granted patent document instead and the claims are taken from it. Reading and scoring both stay inside China.

Attorney advertising. This page describes services of 北京冠和权律师事务所, a law firm licensed in the People’s Republic of China.