> ## Documentation Index
> Fetch the complete documentation index at: https://docs.fetchlake.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Criteria

> Define what gets a paper in or out.

Criteria are the rules you screen against. Huxley uses them in two places: to highlight matching words while you screen, and as the instructions Huxley AI follows.

## Adding a criterion

Each criterion has:

| Field | Options |
| - | - |
| **Type** | **Inclusion** or **Exclusion** |
| **Rule** | **Must Include**: the record must mention one of the keywords. **Must Exclude**: a match rules it out. **Boolean Expression**: a full rule with `AND`, `OR`, `NOT` and brackets. |
| **Keywords** | Comma-separated words or phrases (for Must Include / Must Exclude). |
| **Description** | Plain-English explanation. This is what the AI reads most closely. |

### Boolean example

```text theme={null}
(child* OR adolescen* OR "young people") AND (anxiety OR depress*) NOT (rat OR mice)
```

## Versions

Every time you run Huxley AI, the current criteria are saved as a **version**. Each AI decision is linked to the version used, so if you change your criteria part-way through you can still tell which decisions were made under which rules.

## Writing good criteria

* Write one criterion per idea: population, intervention or exposure, comparison, outcome, study design (PICO/PECO).
* Say what to do with edge cases, for example "conference abstracts: exclude" or "mixed adult and child samples: include if results are reported separately".
* Keep the wording the same as your protocol, so your review can be checked against it.
* If you change criteria after screening has started, re-check the decisions made under the older version.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.