Can AI write a patent application?

What AI does well, and badly, in drafting

The underlying problem is that a patent application is fixed on filing. In the US the description must show the inventor had the invention and enable others to make and use it (35 U.S.C. 112(a)), and at the EPO an application cannot be amended to add subject matter beyond what was filed (Article 123(2) EPC). A missing embodiment cannot be added later, and an invented one, described with confidence, stays in the record. Fluent text makes both mistakes harder to spot, not easier.

A worked example

An inventor describes a water bottle cap with a UV-C LED that sterilizes the water for 60 seconds when the cap is twisted. A capable AI draft will include a clean independent claim to the cap, dependent claims on the switch, the timer and the battery, and a description of the circuit. It may also describe a version that uses a sensor to detect bacterial load. If the inventor never made that work, it should come out. And the draft will not know that UV-C caps with twist activation were already on sale, so claim 1 may read on the prior art. A search, then a practitioner narrowing claim 1 to what is actually new, such as the specific timing control, is what turns the draft into an application worth filing.

What the USPTO says about AI tools

On April 11, 2024 the USPTO published guidance on the use of AI-based tools in practice before the Office (89 FR 25609). It creates no new rules; it explains how the existing ones apply. We found no later replacement as of October 26, 2026. The points that matter for drafting:

• Review is required. By signing, a practitioner or applicant certifies the paper after a reasonable inquiry (37 CFR 11.18). In the USPTO's words, "Simply relying on the accuracy of an AI tool is not a reasonable inquiry." Citations must be checked.
• AI cannot sign. Correspondence signed by an AI tool is not acceptable, and AI systems may not hold a USPTO.gov account.
• Disclosure. There is no general duty to tell the USPTO an AI tool was used. Use must be disclosed where it is material to patentability under the duty of candor (37 CFR 1.56), for example evidence that a named inventor did not significantly contribute.
• Confidentiality. Practitioners must make reasonable efforts to prevent inadvertent or unauthorized disclosure of client information (37 CFR 11.106), which means reading the AI tool's terms, privacy policy and security practices, because some tools retain what users enter.
• Export and foreign filing. AI tools may use servers outside the United States. A foreign filing license does not authorize exporting subject matter abroad to prepare a US application, so a US invention sent to a tool that processes it abroad raises 35 U.S.C. 184 and 37 CFR 5.11 questions.

Inventorship: AI cannot be an inventor

Every major office and court to rule on it has said an inventor must be a person.

• Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. August 5, 2022): the Patent Act requires an inventor to be a natural person. The Supreme Court declined to hear the case on April 24, 2023.
• EPO J 8/20 (December 21, 2021): "A machine is not an inventor within the meaning of the EPC." The designated inventor must be a person with legal capacity (Article 81, Rule 19 EPC).
• [2023] UKSC 49 (December 20, 2023): under the UK Patents Act 1977, an inventor must be a natural person.

The harder question is when a person who used AI is the inventor. In February 2024 the USPTO issued guidance that applied the joint-inventorship Pannu factors to each person working with an AI system. On November 28, 2025 it rescinded that guidance in its entirety (90 FR 54636). The revised guidance says there is no separate standard for AI-assisted inventions: conception is the test, AI systems are tools like lab equipment or software, and the Pannu factors apply only when deciding between several natural persons. An application naming an AI system as an inventor is rejected, and a priority claim to a foreign application naming only an AI as inventor will not be accepted. Articles still citing the February 2024 test are out of date.

The EPO's position

The EPO's Guidelines for Examination (2026) state it plainly: "The parties are responsible for the content of their patent applications and submissions to the EPO", and for complying with the EPC and PCT, regardless of whether a document was prepared with the help of an AI tool. The EPO uses AI in its own work under an internal AI policy. It has not issued a rule requiring applicants to disclose AI use. European patent attorneys are also subject to their professional body's conduct rules, so check with your representative how they use AI on your file.

Confidentiality: questions to ask any AI vendor

An unfiled invention is a secret, and in most countries outside the US, disclosing it before filing can destroy novelty. Before uploading a disclosure to any AI tool, get written answers to these:

• Is my data used to train or improve any model, yours or your model provider's?
• In which countries and regions is it processed and stored?
• How long is it kept, and what happens when I close the account?
• Is there a data processing agreement and a list of sub-processors?
• Who at the vendor can see my documents, and when?

For US-origin inventions, the location question matters for foreign filing licenses and export control. That is a question for your own counsel, not the vendor, but you need the vendor's answer to ask it.

More patent guides

• Patent drawing requirements: USPTO, EPO and India

  37 CFR 1.84, the EPO rules that replaced Rule 46 EPC, and Indian rule 15, compared side by side.

• How to write patent claims: USPTO, EPO and India

  Claim structure and worked examples, then the claim rules and fees at the USPTO, the EPO and the Indian Patent Office, side by side.

• Types of patent searches, and which one you need

  Patentability, freedom-to-operate, invalidity and landscape searches compared by purpose, timing, scope and deliverable.

• Patentability search: how to check novelty before you file

  What counts as prior art under 35 U.S.C. 102 and Art. 54 EPC, the grace period gap, search steps and how to read results.

• Freedom-to-operate search in the US and Europe

  In-force claims only, legal status checks at the USPTO, EPO and UPC, FTO opinions after Halo, and design-arounds.

• Invalidity search: prior art to challenge a patent

  IPR, PGR, EPO opposition and UPC revocation deadlines, prior art dates, standards documents and claim charts.

• Patent landscape analysis, step by step

  Building a CPC and keyword dataset, cleaning families and names, the metrics that matter and the 18-month publication lag.

• Google Patents search guide

  Search operators and fields, Boolean and proximity syntax, non-patent literature, and where Google Patents falls short of the official databases.

• The Cooperative Patent Classification (CPC), explained

  How CPC symbols are built, CPC against IPC, the Y section, how to find the right class and how to search it in each database.

• Patent search Boolean operators and search strings

  Boolean, proximity and truncation syntax compared across Patent Public Search, Espacenet, Google Patents and The Lens, with semantic search for contrast.

• The PCT application: timeline, fees and national phase

  The international phase month by month, receiving Office and ISA fees in CHF, USD and EUR, national phase deadlines by office, and a deadline calculator.

• How much does a patent cost?

  USPTO fees by entity size from filing to the last maintenance fee, what drives professional fees, EPO fees, and an estimator.

• How long does a patent last?

  US utility and design terms, patent term adjustment and extension, terminal disclaimers, maintenance fee deadlines, Europe and India, with an expiry estimator.

All guides, by patent office