Use prior-art search when you want &AI to find patents and published applications that may be relevant to the target patent's claims. This is the main search workflow for invalidity research and reference discovery.

This workflow is available on all plans and costs tokens to run.

Prior-art search now runs inside a Search Agent thread anchored to the patent you are analyzing.

Start on the project's Search page and click New search.

Choose the target patent you want to search against and optionally add a title for the thread. Once the thread opens, ask Search Agent for prior art instead of switching into a separate search-session mode.

A good prior-art prompt usually includes:

You can ask Search Agent to focus on CPC codes, assignees, date ranges, citation paths, specific claim limitations, or combinations of references.

Use follow-up prompts when you want to tighten or widen the search. For example, you can ask for more results near a strong reference, exclude a noisy concept, focus on a narrower date window, or look for patents that combine with an already promising reference.

Keeping the work in one thread makes the search history easier for your team to revisit later.

As the run completes, review the patent finalists in the thread or results panel. Focus on whether the reference looks relevant to the target patent, whether the dates fit the theory you are testing, and whether the result is strong enough to bring into the project.

Approve the references you want to keep, reject the weak ones, and leave others pending if you want to compare them first.

When you have approved the right finalists, use + Add Selected to import them into the project.

Approved prior-art results are added as project documents so you can continue into document review, processing, and charting.

From here, continue refining the same thread, add stronger references into the project, or move into downstream charting and document review.