---
title: Teach
description: Teach puts tests on a page, such as a table cell, a form field or a line, for a person to judge Accurate or Inaccurate. Every reading and verdict is kept as a dated row.
section: Filing and documents
order: 8
---

A test is one value on a page, named by its row and column, or by its label. An agent adds tests with what it read; a person judges them in [Doc Review](/docs/filing/doc-review). A **High** test is one to judge first, with the reason why.

## Judge a test

1. Open the case in Doc Review and choose the **Review** tab. Its **List** has a **Teach** section, page by page.
2. Choose a test. Its page opens with its box highlighted, and its card is in **Next Task**. On the page, point at a test's box to see its card; click the box to keep the card open.
3. Choose **Accurate** or **Inaccurate**. For Inaccurate, type what it **should read** and pick a reason: **Value**, **Box**, **Row or column** or **Other**. A comment is optional.
4. Choose **Submit**. The next test opens, High tests first.

On the page, a test's box shows where it stands: blue waiting, green Accurate, red Inaccurate, amber Changed. A High test's box is heavier and its number filled.

## Every decision stays

Each reading and each verdict is a row under its test, with who made it and when. A new verdict reverses an earlier one, and both stay. When an agent adds a reading that differs from the one you judged, the test is **Changed** and asks again.

Only a person judges. An API key or a connected agent adds tests and readings, never verdicts.

## Add a test of your own

Click a cell, field or line on the page, and **Add as a test** appears beside it. The parse's reading there becomes a test, ready for your verdict.

Or draw one: choose **Draw a test** on a page's ⋯ menu or the toolbar, drag a box around the value, then choose **Add as a test**. The test starts with the page's own text inside the box, and its card opens for your verdict. Nothing is cut or uploaded.

A test you add has no name: it goes by its page and number, such as **p9.4**.

## The report

The Teach section's bar copies or downloads the report: the counts, then each test with its rows, in Markdown. An agent reads it to fix what's wrong. Through the API, see [endpoints](/docs/api/endpoints).

## Example

An agent reads `$1,250.00` for the appraisal fee under **Borrower-Paid · At Closing**. You mark it Inaccurate, **Row or column**, with the comment "it's under Before Closing". The agent adds a new reading under **Before Closing**, and the test turns Changed. You mark it Accurate, and the report counts it as fixed.