# Inspect

> What the factory found in your file, how serious each finding is, and what to do next.

Step 3 on your Factory page. **Inspect** reads your current manuscript and
tells you what won't come through the way you meant: stray styles, headings
the factory can't see, odd spacing, images, tables and notes. It's **free and unlimited** and
never uses a build. A script checks your file against a list of rules. No AI
reads your manuscript.

Choose **Inspect manuscript**, then **Inspect again** after each revision.
The box shows how many findings there are of each kind, with a **learn more**
link to each kind's page. **Open the full report →** shows every finding in
detail.

## How serious is it?

| Label | Meaning |
|---|---|
| **Worth fixing** | Likely to come out wrong in the book. Fix it in your editor and upload again. |
| **Worth a look** | May be fine; check it. |
| **Just noting** | For your information. The build handles it. |
| **Carried through automatically** | The build keeps it as you meant it (your italic and bold, for example). Nothing to do. |

**Nothing Inspect finds blocks a build.** You can always build a free proof
and look. A clean file shows *Things to look at: 0*.

## The full report

The report opens in place of the steps, with the step list still on the
left; **Back to the steps** returns. It starts with one sentence, such as
*2 things are worth fixing before you export a final*, then a table with one
row per kind of finding, grouped by the labels above:

- **What** it is and where (¶ is the paragraph number);
- **What the build does** with it;
- **You**: what to do, often *nothing*.

Click a row to see each place it occurs. From the report you can also:

- **correct as is**: on a *Worth fixing* or *Worth a look* row, or one
  place in it, say the text is meant. It stops being flagged and moves to
  *Carried through*, marked *you said: correct as is*. **undo** puts it
  back. It holds for this upload; a new upload is inspected afresh.
- **fix here**: for stray ">" marks and markers the factory doesn't
  recognize, correct the text without opening your file. The fix is made on
  every build and listed under **Your corrections**, with how many places it
  changes and a **remove** link. **Download the corrected .docx** gives you
  your file with the fixes made; keep editing from that copy, so your file
  and the book stay the same.
- **choose sizes**: on the *Images* row (*print it smaller* on
  low-resolution images), pick a size for each picture; a small page shows
  how it sits. Sizes save at once; **Rebuild proof** shows them in print.
  See [Images](finding-image-inventory.md).

An older report says *an earlier run* and links to the latest.

## The book map

Inspect also shows **how the build reads your file**: what it takes as front
matter, body and back matter, from your Heading 1 titles and where they
fall. Your own title page and contents are dropped, because the factory
generates them from the transmittal. If a chapter lands in the wrong place,
check its heading. The report lists every section with its word count.

## Common findings

- **[Headings without a Heading style](finding-heading-lookalike.md).** A
  line that looks like a chapter title but is styled as body text, so the
  factory can't find it. Give it Heading 1.
- **[Custom styles not in your transmittal](finding-undeclared-custom-style.md).**
  Your file uses a style the factory doesn't know, so its text is set as
  ordinary text. Fine if the style only changed the look; declare it if it
  marks a letter, a poem or a sidebar. Your editor's built-in styles
  (Footnote Text, Hyperlink and the like) are never flagged.
- **[Marked styles](finding-style-marker.md).** The `[[markers]]` it found
  and what each became.
- **[Bold and italic applied by hand](finding-manual-formatting.md).** Kept
  exactly as you typed them. Nothing to do.
- **[Lists](finding-manual-list.md), [breaks](finding-manual-break.md) and
  [spacing](finding-direct-spacing.md) made by hand.** Usually harmless;
  styles are cleaner.
- **[Colored](finding-colored-text.md) or
  [highlighted](finding-highlighted-text.md) text, [unusual
  fonts](finding-unusual-font.md).** Fonts are ignored (the book has its
  own) and highlighting is dropped. Check colored text in a proof.
- **[Images](finding-image-inventory.md).** Every picture: how wide it
  prints on this book's page and its resolution there, with a size menu
  and a small page showing how it sits. Images [too
  low-resolution for print](finding-low-resolution-image.md),
  [floating](finding-floating-image.md) with text wrapped around them, or
  [without alt text](finding-missing-alt-text.md) get their own findings.
- **[Tables](finding-table.md).** Every table, with its header row and
  caption; [tables typed with tabs](finding-tabbed-table.md) and
  [captions on the wrong side](finding-caption-position.md) of their image
  or table.
- **[Footnotes](finding-footnote.md) and [typed note
  numbers](finding-typed-note-number.md).** Real notes are counted; a
  superscript number typed in the text prints with no note.
- **[Emoji in the text](finding-emoji-in-text.md).** ⚠️ ✅ and the like,
  usually notes to yourself. The book's fonts have none.

Every finding has its own page, with why it's flagged and how to fix it in
any editor: see **Inspect findings** in [all help](/help/).

## The counts

Inspect also counts chapters, words and images. They fill in the
transmittal's checklist.

If Inspect **couldn't read the file**, export it again from your editor as a
`.docx` and upload it. Nothing is charged.

---
Source: https://studio.jdbb.net/help/inspect