Case Study: How a City Made Its Tourism Brochure Accessible
It was early on a Monday when the question became urgent at the tourism office of a German municipality — the same question many public bodies have been pushing aside since the introduction of EU public-sector accessibility regulations: what to do about their own PDF publications. The specific trigger was the freshly completed tourism brochure “Discover Our City” — 20 pages, full color, A4, designed by an external designer in Adobe InDesign. Three-column layouts on the event-tips pages, a detailed city map, eighteen photos of landmarks, hotels, and landscape, plus three tables with opening hours and prices. A well-crafted marketing piece. And from an accessibility standpoint, an absolute worst case.
The tourism officer faced several requirements that, at first glance, contradicted each other. As a municipal body, she was obligated to provide accessibility — meaning the document needed to be usable with a screen reader, have a logical tag structure, and offer alt text for the images. At the same time, the design had to be preserved, because the designer had invested several days in the layout. And then there was the printer’s deadline: in five days, the file had to go to print, and before that, the digital version needed to be live on the website. The budget was tight — municipal funds, no room for a four-figure contractor invoice.
The obvious path — and why it wouldn’t work
The first instinct was what many municipalities default to: open Adobe Acrobat Pro and tag the document by hand. Estimated effort for a document of this complexity: ten to fourteen hours at an hourly rate of $75. That’s $750 to $1,050 for a single brochure. Plus the license. And above all: the person with the relevant expertise in-house — that person didn’t exist. The tourism officer is a tourism officer, not an accessibility specialist.
External contracting would have been the alternative. A call to a specialized service provider returned: realistic delivery time of two to three weeks, similar cost. Neither fit the deadline or the budget. A third path was needed.
The attempt with the Hybrid approach
Through a tip from her industry network, the tourism officer came across the AI-assisted Hybrid PDF method. The idea was convincing on paper: the original is left visually untouched — preserved as an unchanged image layer — and an invisible accessibility layer with tags and alt text is laid on top. Sighted users see the unchanged brochure; screen readers access the structured layer. That promised exactly what was needed here: fast, layout-safe, affordable. But the skepticism remained: can automation really deliver accessibility-conformant results, or is this marketing fluff?
The conversion itself took surprisingly little time. Upload of the 12 MB file: thirty seconds. Complexity analysis: another minute or so. The score landed at 8 out of 10 — not a simple case. The system detected the multi-column sections on six pages and automatically routed them into a specialized wrapping procedure. The three tables were processed with automatic header detection. For the eighteen images, the vision-model pass for alt-text generation kicked off in parallel. The city map — a complex vector graphic — the system flagged for an extended long-description.
After about six minutes, the finished Hybrid PDF was in the dashboard. Internal validation reported: tag structure complete, reading order logical, eighteen of eighteen images with alt-text suggestion, table headers correctly marked, document language defined as German, eight automatically generated bookmarks, complete metadata. PDF/UA conformance: passed.
Trust is good, external validation is better
At this point, one could have called it done — but caution was warranted. The tourism office additionally uploaded the document into PAC 2024, the free validator recommended by the PDF Association. Result: passed here too. An additional spot check with the NVDA screen reader confirmed that the read-aloud experience worked — the heading structure was recognized, the multi-column sections were read correctly from column one to column two to column three, the tables were navigable, the images were described.
The contrast with the original was striking. In the original brochure, without a tag tree, without alt text, without marked table headers, without language definition, NVDA had read only fragmented text, ignored the images entirely, and skipped right over the city map. In the Hybrid PDF, everything was conveyed in structured form, with heading-level navigation possible — yet visually the document was absolutely identical to the original. Pixel comparisons showed no visible differences.
What didn’t work automatically
The story would be too good to be true if it didn’t honestly cover what didn’t work perfectly on its own.
First, the long description of the city map. The AI had captured all landmarks correctly and arranged them in geographic order — from north to south. Factually correct, but not very useful for tourists. The designer adjusted the order afterward and sorted by tourist recommendation: first the main attraction, then the nearby points, then the day-trip destinations. Effort: about ten minutes.
Second, a single photo: the mayor at the opening of a hiking trail. The AI had described him as “person in suit at an outdoor event.” Factually accurate, but devoid of context. The designer, with the consent of the person depicted, added the name and the occasion. Effort: two minutes.
Third, the hyperlinks. Several links to the tourism association’s website had generic anchor texts like “learn more” or “click here” — typical WCAG 2.4.4 violations, because when screen-reader users tab through all links in a document, they cannot tell where “click here” actually goes. These were rewritten into clear descriptions: “open the 2026 event calendar,” “view the Bad Belzig hotel list.” Effort: about twenty minutes.
And finally, something PDF/UA does not check at all: color contrast. One section of the brochure used light-gray text on white background — aesthetically pleasing, but just below the WCAG minimum of 4.5:1. The digital version went out as-is because the print deadline was looming; in the next print update, the contrast was adjusted. An important lesson: tag structure alone is not everything. WCAG aspects like color contrast and link-text quality must be checked separately.
In total, manual rework came in at thirty to forty-five minutes — a fraction of the ten to fourteen hours of manual tagging that would otherwise have been required.
The cost calculation
Putting the two paths side by side makes the economic difference tangible:
| Method | Time | Personnel cost | Tool cost | Total cost |
|---|---|---|---|---|
| Manual with Adobe | 10–14 h | $750–1,050 | ~$20 (license pro-rata) | ~$770–1,070 |
| AI-assisted + 30–45 min QA | ~45–60 min | $55–75 | $5–10 (plan pro-rata) | ~$60–85 |
Possible savings on this project: roughly $700 to $1,000 and several hours of work — with comparable end results after manual validation. This is not a theoretical calculation; this is the lived experience of a municipal project.
Feedback from the Field
“We were skeptical whether automatic conversion could really be conformant. The result convinced us — faster, more economical, and the layout looks like the original. Exactly what we needed.”
— Tourism officer, Germany (anonymized)
What this project teaches
For public bodies, the most important takeaway is this: AI-assisted conversion can meet accessibility requirements — provided you don’t naively misunderstand it as full automation, but treat it as a tool that does the bulk of the work and leaves room for targeted manual refinement. This path works even under time pressure, and external contracting can often be avoided, provided there is internal time for validation and small corrections.
For designers and agencies, layout fidelity is the decisive argument. In the Hybrid approach, the original is not reconstructed from tags (with all the layout-drift risk that entails) but preserved as an unchanged image layer. What the designer delivered remains visually exactly as conceived. A half-hour QA round suffices in most cases to make the document delivery-ready.
More generally: multi-column layouts, formerly the stumbling block of many conversion tools, have become manageable with modern routing methods. Vector graphics like maps still require manual long descriptions, because no AI knows the strategic ordering of landmarks for a tourism audience. Photos of people should be supplemented context-specifically; generic descriptions are not enough. And color contrast as well as link-text quality should be checked separately, because PDF/UA does not cover those WCAG requirements.
Worth trying yourself
If your tourism board, your administration, or your agency faces a similar task: try it yourself, ideally with a real, medium-complexity document from your existing inventory. After one run, you’ll have a grounded assessment.
No responses yet