How Phrasemill sources data

Phrasemill is built around evidence authors can inspect. Taste tags are extracted from public agent and agency websites, then linked back to the source text wherever possible.

Primary sources

Our primary source is public website text from agents and agencies: bios, submission pages, FAQs, wishlist pages, and similar material. When a tag appears on a profile, Phrasemill stores the quoted span that supported it and shows a link to the source page when that link can be resolved.

Secondary sources

Some represented-book and author signals use public Open Library metadata, such as book subjects or author information. When evidence comes from Open Library rather than an agent's own page, Phrasemill labels that provenance separately.

How tags are made

The pipeline crawls source pages, extracts taste signals with quoted spans, normalizes those signals to a controlled tag vocabulary, and materializes the current tags on agent and agency profiles. Evidence is loaded separately so you can check the underlying text before deciding whether a match is useful.

What we do not use for tag evidence

Phrasemill does not use interviews, MSWL feeds, deal databases, or deal announcements as taste-tag evidence. If a profile includes an MSWL link, that means the link was found on the agent's site; it does not mean MSWL entries were ingested as tag sources.

Reporting a problem

If something looks wrong, email hello@phrasemill.com. Include the agent, agency, or tag you were looking at and what seemed off.

Phrasemill is a product of AK Foundry Ltd, registered in England and Wales (Company No. 16882461).