Guide
How to tag an untagged PDF
Auto-tag produces a structure tree that is wrong in predictable ways. The order to correct it in, and why that order saves about a third of the time.
About 17 minutes · involved
What auto-tagging gets right and wrong
Acrobat’s Add Tags to Document — or any equivalent — builds a structure tree by analysing layout. It is genuinely good at some things and reliably bad at others, and knowing which is which tells you where to spend your time.
Reliably right: which glyphs form a paragraph, which marked content belongs to which element, which page an element is on, and roughly where tables and images are.
Reliably wrong: reading order in anything but a single column, heading levels, whether a grid is a data table or layout, whether repeated page furniture is content, and every piece of alt text.
So the workflow is: auto-tag to get the associations, then correct the semantics. Do not tag by hand from an empty tree — you would be redoing the part the machine does well.
Before you tag
Check it is not a scan. Select all, copy, paste. Nothing pastes → see the scanned-PDF guide; tagging cannot start until there is text.
Check for an existing broken tree. File → Properties → Description shows “Tagged PDF: Yes/No”. If it says Yes but the document reads badly, you have a bad tree rather than no tree — Acrobat will refuse to auto-tag over it. Delete the existing tree first (Tags panel → select the root → Delete), or work with what is there.
Save a copy. Acrobat’s Tags panel has no multi-level undo for drag operations. You will want the copy.
Step 1 — Auto-tag
All tools → Prepare for accessibility → Autotag Document.
On a 40-page report this takes a minute or two and produces a tree with a few hundred elements. Do not look at it yet.
Step 2 — Artifact the page furniture
Do this first, before anything else, and it is the step people skip.
Every running header, footer, page number, decorative rule and background element is currently a tagged paragraph sitting in the reading order. On a 40-page document that is 80–120 elements you would otherwise be dragging past for the rest of the job.
All tools → Prepare for accessibility → Fix reading order, marquee each region, click Background/Artifact. See the artifacts guide for what qualifies and what does not — a logo in a header is decoration on page 7 and the issuing organisation on page 1.
Removing them now makes every subsequent step shorter.
Step 3 — Fix the reading order
Now the tree contains only content, and the reorder is tractable.
Open the Tags panel and the page side by side. Click each element to see its region highlighted — that is the only reliable way to know what you are looking at. Drag runs, not individual elements, wherever you can.
Multi-column pages are where the work is. For a right-to-left document the correct order starts with the right column, and no auto-tagger does that — see TG-RTL-004. Full detail in the reading-order guide.
Do this before alt text and before headings. Both attach to elements that this step moves.
Step 4 — Correct the heading levels
The auto-tagger guesses headings from type size, so it produces a plausible-looking outline that is usually wrong about depth: section headings and subsection headings that share a size become the same level, and a large pull-quote becomes an H1.
Retype the elements in the Tags panel, then read the resulting outline as an outline. If it does not describe the document, it is still wrong. The headings guide covers the three failure shapes.
Step 5 — Rebuild the tables
Auto-tagged tables are Table → TR → TD, with no TH anywhere, because the tagger cannot tell a header from a value.
For each table: decide whether it is data or layout. A layout table should not be a table at all — retype it to a Div or ungroup it. A data table needs TH cells with /Scope, and if it has merged cells or two header rows it needs Headers/ID references. The table guide has the grammar.
Tables are the slowest part of most documents. Budget accordingly.
Step 6 — Lists
Bulleted text auto-tags as a run of paragraphs starting with bullet characters. It needs to become L → LI → (Lbl, LBody).
In the Tags panel you can select a run of paragraphs and use Create Tag from Selection, but nesting the LI and LBody levels is manual. For a document with many lists this is worth doing in the source and re-exporting if the source exists at all.
Step 7 — Figures and alt text
Every image is now a Figure with no alt text. Two decisions per image, in this order:
- Is it decorative? If yes, change the tag to Artifact. Do not give it empty alt text — an empty
Figurestill occupies a position in the reading order. - If not, what does it say? Write alt text that carries the information the image carries. A chart’s alt text is its finding, not “bar chart”. See writing alt text.
Alt text comes after the reorder because the reorder moves the figures, and after the artifacting because artifacting removes some of them from the list entirely.
Step 8 — Links, forms, and the document level
Links need /Link structure elements, not just annotations — see the links guide. Form fields need tooltips — see the forms guide.
Then the document level, which takes two minutes and is the highest-value part of the whole job: the title, the display-title flag and the language. The title guide.
Step 9 — Verify
Two checkers, then a person. How to check free, and a script for the human pass.
The human pass is not optional here. Everything in steps 3 to 7 is a judgement, and a validator confirms none of it.
What this actually costs
For a text-heavy 40-page report with a few tables and images, working in this order: three to five hours for someone who has done it before.
Out of order — alt text before reordering, headings before artifacting — the same document runs six to eight, because several steps get done twice. That gap is the entire reason the order above is worth following.
Frequently asked questions
Should I auto-tag, or tag by hand?
Auto-tag, then correct. Building a structure tree from nothing in the Tags panel is many hours of work for a document of any length, and the auto-tagger gets the tedious parts — paragraph boundaries, marked-content association, page association — right. It gets the interesting parts wrong, and those are the parts you would want to do yourself anyway.
Acrobat's Add Tags produced a mess. Should I delete it and start again?
Almost never. A mess with the right marked-content associations is much closer to done than an empty tree. The exception is a document where the auto-tagger has scrambled a complex layout so badly that the associations themselves are wrong — a heavily floated magazine spread, say. Then re-tagging one page by hand can be faster than untangling it.
Why does the order of corrections matter?
Because they depend on each other. Artifacting page furniture removes elements the reorder would otherwise have to move. Fixing the reading order moves the figures the alt text is attached to. Doing it in the wrong order means doing several steps twice, which on a real document is roughly a third of the total time.
Can Taggart do the tagging?
It creates P and Figure elements in painting order and says exactly that in the change log — painting order is not reading order, so it is a first pass, not a finished job. The reading-order editor is where the actual work happens, and that is deliberate: an auto-tagger that presented its output as finished would be lying about the hardest part of the task.
Related failure conditions
- 01-005 Content is neither marked as Artifact nor tagged as real content Every text-showing, image-painting and path-painting operator on a page must sit either inside a marked-content sequence that maps to the structure tree, or inside an artifact sequence. Content that is in neither is orphaned.
- 09-001 Tags are not in logical reading order The order of elements in the structure tree does not match the order a human would read the page. This is the single most common substantive defect in tagged PDFs and the one Acrobat's own tooling handles worst.
- 14-001 Headings are not tagged Text that is visually a heading — larger, bolder, isolated — is tagged as a paragraph. The document has no navigable outline.
- 13-001 Graphics objects are not tagged with a Figure tag An image or vector drawing that carries meaning is not inside a Figure element. Either it is untagged, or it has been swept into a paragraph.
- 18-001 Headers and footers are not marked as pagination artifacts Running heads, footers and page numbers are tagged as real content, so they repeat in the reading order on every page.
Check your own file. Taggart's validator is free and unlimited — no page cap, no watermark, no account.
Last updated .