Figure Annotation for Claim Charts: Sharper Infringement Analysis

Most discussions of patent infringement analysis stop at detection: finding where an accused product reads on the claims. But anyone who has built an evidence-of-use chart by hand knows detection is only the start. The chart still has to make that read legible to someone else. A limitation maps to a region of a product figure, a teardown photo, or a frame from a demo video, and getting that onto the page so a reader sees it is what the deadline never leaves room for.

So the figure gets shortchanged. You end up with an arrow pointing at the whole image, or no annotation at all. The citation is sound, but the chart records it without explaining it, and the reader has to reconstruct the argument the chart was supposed to hand them.

Compounding this, the same chart travels to very different audiences, and each one wants something different from the figure.

A claim chart has at least two jobs

What the annotation needs to do depends on what the chart is for at that moment in the case.

The teaching version goes to a judge, a jury consultant, an expert, or a litigation funder. These are readers who flip through and look at pictures, so specific annotations pay for themselves: a box on the accused component, an arrow to the contact surface, a label tying the image back to the limitation language. You're teaching the read to someone who hasn't lived inside the patent for months.

The pleading-stage version is the opposite. It goes out with a complaint or early contentions, where a general arrow, or none, is often what you want. It gives notice while leaving room to refine the theory as the case develops. Most matters use both. You build the teaching chart first to prove the read to yourself, file the stripped-down version, then bring the teaching chart back at expert reports.

The teaching version is where careful annotation earns its keep, and it's also the version that gets cut, because detailed markup is the part of the job that doesn't scale by hand.

What we built

&AI's figure annotation closes that gap. When the charting engine maps a limitation to a figure, whether a patent drawing, a product spec image, a screenshot, or a video frame, it produces the annotation as part of the output rather than leaving it as a later chore.

The engine takes three inputs: the figure, the limitation, and the citation text. From those it generates annotation shapes, including boxes, arrows, highlights, and labels, in normalized coordinates so they render cleanly at any size. Crucially, the label text comes from the record itself, drawn from the figure's own text, the citation, and the surrounding context, not from the model's guess about what the picture shows. A well-annotated chart becomes the default output, not the thing you manage to produce when there's time left over.

Annotation density is also a setting. Export controls govern how heavily figures are labeled, how they're formatted and cited, and where they sit in the document. The teaching chart and the pleading chart come from the same evidence, with no re-annotation in between.

The impact

For litigators, the explanatory chart no longer competes with the deadline. The version you'd want in front of an expert or a jury is what the engine produces by default, and you can still dial it back for a filing. For funders weighing a portfolio, the figures explain themselves, so you can move across patents and accused products and actually see each theory rather than taking the citation on faith.

Annotation is what separates a chart that records a read from one that makes it. It gets skipped because it costs time, not because anyone doubts it's worth doing. Take that cost out, and you stop under-explaining the strongest part of your case.

&AI turns source-grounded figure annotation into trial-ready work product. Explore claim charts, prior art search, and infringement detection, see pricing, or book a demo to try it on your own matter.