How AI-assisted content structuring is eliminating InDesign production bottlenecks

TL;DR

A textbook chapter can look perfect in InDesign and still break the second you turn it into an ebook. On the page, the title is big and bold, the headings stand out, the sidebar sits in its own little box. But export it to EPUB and all of that disappears.

The titles and headings turn into plain text, the sidebar gets dumped into the middle of a paragraph, and captions stick to whatever was above them. It looked organized, but nothing underneath was actually telling the computer what each part was.

That’s the real bottleneck in publishing to multiple formats. Fixing one messy file can take longer than building the chapter did, and you have to do it over again for every format you produce. Below: why InDesign causes this, what the cleanup really costs, and how AI fixes it.

InDesign is made for layout, not structure. A file can look right on the page and still have nothing underneath that tells a digital format what each part is, no headings, lists, tables, or reading order, which is exactly what those formats need to work.

So you end up cleaning up the same things over and over. Every time content goes from a print-ready InDesign file into EPUB, accessible HTML, or an interactive textbook, someone has to go back in and re-tag or rebuild the structure by hand.

That step is one of the biggest hidden costs in multi-format publishing. You pay it again for every title, every edition update, and every new format you ship to.

AI content structuring reads the content coming out of InDesign, works out which parts are headings, body text, captions, tables, and lists, and tags them consistently without anyone doing it manually.

KITABOO Fluid 360 runs this structuring inside a larger content transformation pipeline, turning layout-first files into structured content that’s ready for accessible, multi-format output.

Table of contents

1. Why InDesign creates a structural bottleneck
2. The hidden cost of manual re-structuring
3. What AI-assisted content structuring actually does
4. Where AI structuring fits in the production pipeline
5. The AI content structuring readiness checklist
6. How KITABOO Fluid 360 removes the structuring bottleneck
7. FAQs

InDesign is built around the page. Text frames, image placement, type styling, columns, all of it is set visually and tuned for how a layout reads in print or a fixed-layout PDF. For that job it’s the right tool. The designer decides where everything goes, and the output copies those decisions exactly.

Digital formats work the other way. Reflowable EPUB, accessible HTML, and interactive web content run on meaning, not position: a clear order of headings, paragraphs, lists, and tables, plus a reading order that holds no matter how the page gets rearranged. Each part has to be labeled for what it is. Styling alone doesn’t carry that.

This is where the two split. A heading in InDesign can look exactly like a heading, bigger font, bold, the right spacing, and still carry no tag saying it’s one. To a digital reader it’s just a large, bold paragraph. It looks right and means nothing.

Textbook layouts make it worse. Multi-column pages, sidebars, pull quotes, and call-out boxes all create confusion about reading order once the content leaves the page. Take a two-column spread with a tinted sidebar down the right margin. A print reader knows to finish the left column, move to the right, and treat the sidebar as a side note. The export engine has no such judgment. It might mix the columns together, drop the sidebar into the middle of a sentence, or stick it ahead of the section it belongs to. For someone using a screen reader, that jumbled output can make the content impossible to follow.

The outcome is predictable. An InDesign to EPUB conversion, or an export to HTML, usually gives you a flat or scrambled file. Headings get tagged by hand, reading order gets fixed page by page, and body text gets pulled back apart from captions and sidebars. None of that is writing. It’s rebuilding a structure the layout always implied but never recorded, and that’s the manual step AI-assisted content transformation is built to remove.

This cost is easy to miss because it never shows up as a line item. It’s buried inside conversion timelines and just gets treated as a normal part of digital production. At catalog scale, it’s often the thing actually holding everything up.

The work repeats by its very nature. The same fixes come back chapter after chapter, title after title: tag the headings, pull out the sidebars, rebuild the tables, fix the reading order. Because each InDesign file is built on its own, none of that work carries over. A team that structured 200 chapters last year structures 200 more this year, even when half of them are revised editions of books it already did.

It also lands on the wrong people. Re-structuring usually falls to production or conversion teams who didn’t write the content, so they end up guessing. Is a short bold line a heading or a lead-in? Is a boxed bit a sidebar, a definition, or a pull quote? Different people answer differently, and a catalog that was supposed to follow one standard ends up all over the place.

The time adds up faster than it looks. Say a complex textbook chapter takes two to four hours to re-tag and reorder by hand, headings, sidebars, tables, reading order, and the average title runs twenty to thirty chapters. That is a week or more of conversion work per book, before anyone checks accessibility. Run that across a catalog of a few hundred titles a year and the re-structuring step alone can absorb several full-time roles. These numbers are illustrative, not a benchmark, but any production lead can plug in their own and land on the same conclusion: the cost is real, it is recurring, and it never appears as its own line in a budget.

Every revision starts the loop again. When a print edition changes, the digital version usually gets rebuilt from the updated file instead of patched, so a new edition with three changed chapters can trigger a full pass, because there’s no stable structured source to update. Reliable content structuring automation is what breaks that loop.

Accessibility raises the stakes. WCAG compliance needs a correct heading hierarchy, alt text, table headers, and a defined reading order, and InDesign exports don’t preserve any of that reliably. The step that makes content readable in EPUB is the same step that makes it compliant, so you can’t skip it or push it off.

For publishers running big catalogs, the limit on how fast digital products ship is rarely the writing. It’s the conversion from print-ready to digital-ready, and re-structuring is the heaviest part of it.

This is where AI content structuring for publishers earns its place. It reads content exported from InDesign or a similar layout tool and gives each part a role: heading level, body paragraph, caption, table, list, or sidebar. It works this out by reading patterns in the formatting, position, and type of content, instead of relying on tags the original file never had.

The way it does this is by inference. Where InDesign gives you styling but no label, the styling becomes the clue. A short, bold line in a larger size sitting right above body text is almost always a heading, so the system tags it at the right level and recovers the structure the layout only hinted at. The same logic splits captions from body text, list items from paragraphs, and table cells from the copy around them.

Reading order is the harder part, and the main reason any of this matters. Across multi-column pages, sidebars, and pull quotes, the system figures out a logical sequence for reflowable and accessible output, a call that used to be made by hand on every tricky page. Instead of a reviewer deciding that the sidebar follows the section it supports, the process suggests that order and puts it up for a quick confirmation.

The payoff is consistency at scale. One structuring logic runs across the whole catalog, so the drift you get from different people making different calls mostly goes away. A heading reads the same way in chapter 1 and chapter 40, in book 1 and book 50, and that’s what makes content structuring automation dependable across a long list.

The end result is single-source output. Once content carries consistent structure, that structured content for digital publishing can become reflowable EPUB, fixed-layout EPUB, accessible HTML, and interactive textbook formats from one base. The structuring happens once, not once per format. KITABOO Fluid 360 covers how this fits a full transformation pipeline.

It helps to see where this step sits, because the point isn’t to replace any part of the existing workflow. Content structuring automation only takes over the one part that currently runs on manual labor.

Stage 1, authoring and design.
Content gets created and laid out in InDesign exactly as it always has, for print or fixed-layout PDF. Designers keep their tools and their process. Nothing about the front of the workflow changes.

Stage 2, export and structuring.
Instead of a manual conversion step, the exported content runs through AI-assisted structuring. This is where the semantic tagging gets applied, reading order gets sorted out, and anything that needs a human eye gets flagged.

Stage 3, human review of flagged elements.
Good structuring doesn’t quietly guess on the unclear cases. A styled element that could be a heading or just a decorative callout should be raised for confirmation, not tagged on a coin flip. This keeps quality control in the loop while taking away the bulk re-tagging that used to surround it. The reviewer confirms the edge cases instead of handling every single element.

Stage 4, multi-format output.
From the structured content, the team produces reflowable EPUB, accessible HTML, fixed-layout formats, and interactive textbook content without redoing the structuring for each one.

Looked at this way, AI-assisted structuring is a layer between authoring and output. It leaves both the designer and the digital product where they are and supplies the automation step that today eats the most manual hours.

• Use the checklist below to test a workflow you already have, or a tool you’re sizing up. Every item is a yes or no. A tool that can’t answer yes to most of them is just moving the bottleneck, not getting rid of it.
• Automatic semantic tagging without re-tagging every title. The tool should apply heading, paragraph, list, and table tags on export, so teams aren’t hand-tagging the same structures again on every book.
• Reading-order resolution for complex layouts. It should put multi-column pages, sidebars, and pull quotes into a logical flow, so reflowable and accessible output reads correctly instead of coming out scrambled.
• Ambiguous elements flagged for review. When something could be a heading or a styled callout, the tool should raise it for a person to decide, rather than guess quietly and bury the error in the output.
• One structured source, multiple outputs. The same structured content should produce reflowable EPUB, fixed-layout EPUB, accessible HTML, and interactive formats without rebuilding the structure for each one.
• WCAG-ready output. The result should meet heading hierarchy, table header, and alt text requirements, so accessibility is built in rather than fixed in a separate pass later.
• Updates without a full re-structuring cycle. When a print edition changes, the digital version should update against just those changes, not kick off a full rebuild that repeats work already done.

See how Fluid 360 applies AI-assisted structuring to your existing content. Request a Demo.

Fluid 360 is built to produce structured, accessible, multi-format-ready content without a manual re-tagging pass per title. Its AI-assisted content transformation takes existing publisher content in and returns structured content ready for digital output.

It works on a big catalog because it treats every file the same way. Your InDesign files run through it, and the tagging gets done one consistent way, not differently by each person on each book. That’s the difference between running a catalog and running a hundred separate little projects. You can see how that goes in KITABOO’s publisher case studies.

It handles accessibility and other languages in the same pass, not later. The content comes out with the headings, table headers, and reading order WCAG 2.2 needs, and it can do more than one language at once.

After that, it’s ready for whatever you publish: reflowable and fixed-layout EPUB, accessible HTML, and interactive books in KITABOO. You handle the path from InDesign to digital textbook once, and that one pass covers every format.

It is built for K12, higher ed, and trade publishers who already work in InDesign and want the conversion headache gone without dropping the tools their teams already use.

Request a Demo