FAQs: from content design sin to AI citation strategy

For more than a decade, FAQs have been treated as a red flag in content design. The logic was simple: if you need an FAQ page, your information architecture has already failed.

I still think that instinct was right. But AI-powered search has changed the job. The format we dismissed as lazy can now be one of the easiest ways to get cited, surfaced and trusted, as long as we do it with the same discipline we apply everywhere else.

Here’s what’s changed, and how to use FAQs without undoing everything good content design taught us.

Why FAQs became “bad practice”

In 2013, Government Digital Service published FAQs:why we don’t have them. Sarah Richards argued that FAQs are convenient for writers but more work for readers, because questions take longer to scan than clear, task-led headings.

That critique stuck because most FAQ pages really were (and still are):

dumping grounds for content that should live elsewhere
invented questions nobody asks
marketing copy dressed up as help content

And the deeper point was more important than the format: FAQs often appear because users cannot find information in the places it should already be. That’s not a reason to add an FAQ. It’s a reason to fix the structure.

A List Apart took the same line in 2018, arguing for “purposeful information”, content that exists in the right place, at the right time, in a format that supports both user needs and organisational goals.

That thinking shaped a generation of content designers. FAQs became shorthand for lazy strategy.

What changed: AI search rewards Q and A

Search is no longer just links. Google’s AI experiences and tools like ChatGPT-style answer engines increasingly synthesise content and present the answer directly.

In that environment, being clear is not just a usability win. It’s a visibility strategy.

The research behind “Generative Engine Optimisation” found that small, deliberate changes to content can improve visibility in generative responses by up to 40%, particularly where content includes citations, quotations and statistics.

That matters because Q and A is easy to extract. A system can lift a clean question and a direct answer far more reliably than it can pull the right sentence out of a flowing narrative paragraph.

The critics were right, but only about the lazy version

The anti-FAQ arguments still apply when:

you publish one catch-all “everything” FAQ page
you duplicate answers across multiple pages
you write questions in internal language, not the user’s
you use FAQs to patch broken journeys instead of fixing them

But “never use FAQs” misses something practical: format affects discoverability as well as usability. A question-led heading plus a direct answer is both easy to scan and easy to cite.

Baymard’s e-commerce research is a good example of the nuance here. Their usability testing found the best-performing approach was a hybrid, combining site-authored FAQs with user-generated Q and A. They also noted this was only used by 28% of benchmarked e-commerce sites, which is a useful indicator of the opportunity when you do it well.

The new approach: strategic FAQs, not an FAQ page

Kill the standalone FAQ page

A single page that tries to answer everything is still poor content design, and it’s a weak signal for AI citation because the topical focus is diluted.

Add contextual FAQ sections to relevant pages

Put small, focused FAQ blocks on the pages where people actually hesitate or drop out: product or service pages, application guidance, fees and funding, eligibility, timelines.

This keeps information where it’s needed, while still giving AI systems clean, extractable answers.

Use real questions, in real phrasing

Start with evidence, not opinions:

Search Console queries (filter for what, how, when, where, why)
enquiry logs and support tickets
onsite search terms
“People also ask” patterns for your topics

Then use the exact language your audience uses. Not your internal labels. Not your marketing line.

Structure for extractability

Each item should stand on its own:

Question as a heading (H3/H4, written as a real question)
Direct answer immediately below (2 to 4 sentences that fully answer it)
Then detail (links, caveats, eligibility rules, dates)

If the truth is “it depends”, don’t lead with that. Give the most common answer first, then explain what it depends on.

Make it trustworthy, not just extractable

AI systems do not just look for answers. They look for answers they can trust.

Simple habits help:

add dates (“Fees for 2026–27…”, “Last updated January 2026”)
name the authority where relevant (“UKVI guidance…”, “NHS advice…”)
include numbers carefully and consistently (and link to the source page)
avoid vague claims and superlatives

That last point matters. Over-selling in an FAQ doesn’t just feel wrong. It undermines everything else on the page.

Make it maintainable, or don’t do it

Scaling FAQs across a site can create content debt if you don’t decide who owns them.

A basic governance model is enough:

one owner per answer
one “source of truth” page for policy content (fees, entry requirements, refunds)
review dates, especially for anything time-bound
a clear rule for when to delete or merge questions

Contextual FAQs should often summarise and point back to the definitive guidance, not become a second competing version of it.

Make it easy to fetch and parse

This part is boring, but it’s the difference between “nice idea” and “works in the real world”.

make sure the Q and A is in the HTML and not hidden behind heavy scripts
don’t block key pages behind logins or accidental noindex rules
keep accessibility front and centre, especially if you use accordions

If a user can’t access it easily, an answer engine probably can’t either.

FAQ schema: not dead, just demoted

In August 2023, Google said FAQ rich results would largely be limited to well-known government and health sites, and not shown regularly for most others.

That doesn’t mean structured data is pointless. It means you should treat it as support, not a shortcut. Use schema where it fits, but focus on what it’s always been about: clear structure, clear answers, clear signals of credibility.

How to implement without overthinking it

A pattern that works:

5 to 10 questions per page, max
group by theme if needed (Fees, Eligibility, How to apply)
answers kept tight (roughly 50 to 120 words, then link)
add “Last updated” on the block for anything time-sensitive
remove questions once they stop being frequently asked

How you’ll know it’s working

If you can’t measure it, it becomes another content trend.

A sensible starting set:

growth in impressions for question-led queries in Search Console
increased visibility in “People also ask” or featured snippets (where relevant)
reduced repeat enquiries on the same topics
better conversion through key steps (applications, bookings, sign-ups)
cleaner onsite search patterns (fewer “basic info” queries)

The content design community was right: lazy FAQ pages are a symptom of poor information architecture.

But AI search changes things. We now have to balance two realities: content has to be easy to use and easy to cite.

Strategic FAQs, placed in context, based on real questions, written answer-first, and maintained properly, aren’t a sin. They’re one of the simplest ways to make your expertise legible to both people and machines.

And that’s increasingly the same job