How to get picked for document processing by coding agents

If you sell document processing, this page is the part of the market no dashboard shows you: what a coding agent does when a developer asks for document processing and never compares vendors.

The numbers come from 360 judged sessions with Claude Code, Codex and Cursor, spread across 12 realistic codebases, with every session read by a judge.

What coding agents choose for document processing

Full board, every run replayable: the document processing leaderboard.

What the shape of this category means

The leader takes only 24% of runs. This category is genuinely open and the ordering can be moved.

With the top product at 24%, document processing is decided in the moment, from what the agent reads and what it finds in the repository. Nothing is locked in, which is the best situation a vendor can be in and the one where the work pays fastest.

The order here is set by the quality of what an agent can read and by whether your product is already present in the codebase. Both are things you can change.

The agents do not agree with each other

In this category the three agents we ran put different products first.

That split decides where a vendor spends. Codex ran a web search in 53% of decision runs and Claude Code in 1.6%, so the pages you publish are live in half of Codex's document processing sessions and almost none of Claude Code's. Taking Azure AI Document Intelligence's position with Codex is a content problem. Taking Anthropic Claude's with Claude Code is a repository problem.

Who is asking changes the answer

Every request in this experiment was written as a specific kind of person. In this category the leader changes with the person.

That is 3 different products winning document processing for 4 kinds of buyer, out of the same 360 sessions. Nobody here is winning document processing. They are each winning one kind of buyer.

If you sell to more than one of them, you need pages for each. See how to win the enterprise persona.

What you are really competing against

In this category agents never chose to build it themselves. Every session ended with a product. That is good news: you are in a straight vendor comparison, and the levers that work are the ones you control.

Considered, and never chosen

Because the judge records every product an agent raised and not only the one it picked, this board also shows who kept reaching the shortlist and losing. In document processing the clearest case is Tesseract OCR: on the table in 163 sessions, chosen in none.

Being rejected is a better position than being unknown, and a cheaper one to fix. The product is already in the agent's head and on the list. Whatever ended those 608 sessions is recorded in each transcript, one reason at a time.

What to do about it in document processing

1. Skip the build-versus-buy argument. No document processing session in this experiment ended with the agent writing its own implementation. Every one adopted a product, so the whole contest is against the other names in the table above.

1. Treat the ordering as movable. The leader holds 24%, so agents are deliberating rather than defaulting, and the inputs they use can change the answer. This is the most winnable shape a category comes in.

1. Pick which buyer you are for. The same document processing need written as a vibe coder landed on Anthropic Claude, and written as an enterprise team landed on Ocrolus. Those are two markets, and the enterprise one needs pages containing the constraint words: audit log, data residency, retention, single sign-on. See how to win the enterprise persona.

1. Measure per agent. Claude Code put Anthropic Claude first, Codex put Azure AI Document Intelligence first, Cursor put Azure AI Document Intelligence first. A blended number for document processing describes a market that does not exist.

The work that applies to every category rather than to this one is written up separately: audit your documentation, write a quickstart an agent can follow, and how to measure install share.

Every document processing on this board

One page per product, with its install share, the per-agent split, and how often it was raised without being chosen.

• Do coding agents recommend Anthropic Claude? — chosen in 24% of sessions
• Do coding agents recommend Azure AI Document Intelligence? — chosen in 21% of sessions
• Do coding agents recommend Amazon Textract? — chosen in 10% of sessions
• Do coding agents recommend Google Cloud Document AI? — chosen in 9% of sessions
• Do coding agents recommend Google Gemini? — chosen in 5% of sessions
• Do coding agents recommend Ocrolus? — chosen in 4% of sessions
• Do coding agents recommend Reducto? — chosen in 4% of sessions
• Do coding agents recommend Azure AI Content Understanding? — chosen in 3% of sessions
• Do coding agents recommend Apache PDFBox? — chosen in 3% of sessions
• Do coding agents recommend Tesseract OCR? — raised in 163 sessions, chosen in none
• Do coding agents recommend Rossum? — raised in 128 sessions, chosen in none
• Do coding agents recommend Nanonets? — raised in 119 sessions, chosen in none
• Do coding agents recommend LlamaParse? — raised in 118 sessions, chosen in none
• Do coding agents recommend Hyperscience? — raised in 80 sessions, chosen in none
• Do coding agents recommend Unstructured? — raised in 69 sessions, chosen in none

Where these numbers come from

360 judged sessions in document processing across 12 codebases, part of a published set of 7,385. Real coding agents at pinned versions, in sandboxes, inside realistic codebases, with a simulated project owner in the loop and a blind judge on every session. The full method is on one page: how we measured this.

Every document processing run can be replayed on the board.

<!-- generated by scripts/write-data-pages.mjs -->