About Fishypedia

I'm Nicolas Sartor, and I built Fishypedia because checking whether a lake had been stocked meant hunting down a state agency report, often a PDF or a table buried three clicks deep, and then guessing what it meant for this weekend. So I built the thing I wanted to exist: one page per lake that shows what was stocked, when, and how to fish it.

The problem

Stocking data is some of the most useful fishing information there is, and one of the hardest to actually use:

Fishypedia does that work once: it collects every state's reports into one format, keeps the full history, matches each report to the right lake, and puts the stocking record next to what you need to know to fish there.

What you get here

Where the data comes from

Stocking records come directly from each state's wildlife agency, not from a third-party aggregator: completed-stocking reports where the agency publishes them, and the agency's own stocking plan where it publishes only a plan (labelled as a plan). Lake facts and fishing notes are drafted from the agencies and other named, linked sources. Pages cite their sources and the date each was checked, and when no source says something, the page leaves it out rather than guessing. The methodology page explains what is checked and what is not.

Who's behind this

Fishypedia is published by Nicolas Sartor, who runs it on the side with no team and no editorial board. I do not write the lake, species, or guide text by hand: the data is sourced from agencies and checked by software where it can be, and AI drafts the readable text from the cited sources. The next section explains exactly how. Fishypedia is one of a group of sites I publish and maintain under the Wild Atlas Media network, all built on the same idea: take the federal and state data that's already public, put it in one place, and make it easy to use.

Sister sites: Public Lands DB for who manages the land and its camping rules, closures, and fire restrictions, CampgroundsDB for public campgrounds and how to book them, and ReservoirFinder for reservoir water levels against full pool. Boat ramps and swimming holes are next.

How this content is made

  1. The data is sourced. Stocking records are pulled from state wildlife agency reports into a database, each with the source it came from and the date we pulled it. Lake facts (location, access, species found) come from named public sources and are stored with a source link and the date it was read.
  2. We check it. When a short fishing note is drafted, an automated check requires each claim to match a verbatim quote from a fetched page. A second check compares that page's title, coordinates, county and water type with the lake (a pond is not a lake, a creek is not a river) and rejects pages about a different water; it can still miss two waters that share a name. Each lake's map pin is also checked against its state and, where an agency publishes a county, against that county; a pin we cannot verify is not drawn, and the page shows no map, elevation or nearby lakes for it. On every build another check rejects notes that are about a different water, restate a fishing rule, carry a health, water-quality or safety claim (advisories, bacteria, algae, closures), present a one-off event, plan or ongoing operation as a standing fact, say a species is stocked when the site's own stocking record does not, or say nothing about fishing. For lake overviews, the checker compares numbers and years with the cited pages it can read. Some cited pages, such as PDFs and sites that block automated readers, cannot be read by the checker, so for those the link is there for you to check. We add spot checks on top, and text that fails a check is removed, not guessed.
  3. A language model writes the readable text. The short fishing notes on lake pages are drafted by an AI model from fetched, cited pages only, inside a strict schema where each claim is tied to a verbatim quote. Lake overviews, spots and tactics, species profiles, and guides are drafted by AI agents from the sources they cite, against a typed schema in which each field carries its own sources; those fields are not quote-tied, and audit scripts flag missing sources instead. Where you see such text, it is labelled: "Drafted by AI from the cited sources and checked before publishing."
  4. We keep auditing. The build runs automated content audits, and we audit published pages against their cited sources from time to time. When we find an error we fix the cause, not just the one page, and remove text we cannot support.

Some sentences are not drafted by a model at all: counts, dates, "last stocked" lines, and the lake, county, and state summaries are generated by software from the records, not written by hand. Illustrations without a photo credit are AI-generated and labelled as such.

The same process applies across the Outdoor Content Network: shared refresh cadence, indexability gates, and reviewer steps.

You can check the claims yourself: each AI-drafted passage sits next to links to the sources it was drafted from and the date we checked them, and every lake page lists all of its sources. If something is wrong, tell us and we will fix it.

Tell me what's missing or wrong

If a lake near you isn't accurate, a stocking report is missing, or there's a feature you wish this had, I want to hear it. Email nicolas.sartor@wildatlasmedia.com.