Usually because it reads like a brochure instead of an answer. Most service pages open with a slogan, then a promise, then a form. There is nothing on the page that directly answers "who does this, for whom, at what price, and how long does it take". An AI model has nothing quotable to pull.
We see this constantly. A studio or agency ranks fine for its brand name, but when a buyer asks ChatGPT or Google AI Mode for a recommendation, the page never comes up. The service is real. The work is good. The page just does not contain the facts the question needs.
The fix is less exotic than most AEO advice suggests. It is mostly writing plainly, answering the real question near the top, and backing it with structured data and proof.
It means making the page easy to extract a factual answer from. AI systems read your page, pull the sentences that answer the query, and cite the source. So the work is writing clear, self-contained statements about what you do, plus the normal technical SEO that gets the page indexed in the first place.
There is a lot of noise around this. Vendors sell AEO packages, GEO audits, and AI-specific file formats. Most of it dresses up ordinary content work in new vocabulary. The underlying job has not changed much. It has just become less forgiving of vague copy.
What has genuinely changed is the format. A service page that buries its answer in paragraph six is now competing with pages that answer in paragraph one. Extraction rewards the direct page.
No, and Google says so directly. Its documentation on AI features states there are "no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary". The bar is being indexed and eligible to show with a snippet. That is it.
Google goes further and rules out the file formats people keep selling. In the same documentation it says "you do not need to create new machine readable files, AI text files, or markup to appear in these features". If a consultant tells you their proprietary AI file is the unlock, that claim contradicts the platform own guidance.
Google does describe how the surface works. It says AI features use a "query fan-out" technique that can surface "a wider and more diverse set of helpful links" than a standard results page. That is the real opportunity. One buyer question becomes many sub-queries, and a page that answers several of them can appear where it never ranked before.
Google also lists what it does recommend, and none of it is new. Allow crawling in robots.txt. Use internal links so content is discoverable. Keep important content in textual form. Match your structured data to the visible text on the page. We walk through the practical version of that in our guide to optimizing for Google AI Overviews.
Put a complete answer in the first 50 words of each section. Name the service, the audience, the deliverable, and the constraint in plain sentences. Avoid pronouns that need earlier context. Each paragraph should make sense if someone lifts it out and shows it to a stranger.
The test we use is simple. Copy any single paragraph from the page, paste it somewhere with no surrounding text, and read it. If it still tells you something true and specific, it is extractable. If it reads like "we bring your vision to life with a bespoke approach", it tells a reader nothing and gives a model nothing to cite.
Specificity beats polish here. "We build Webflow marketing sites for B2B software companies, usually 15 to 30 pages, in four to eight weeks" is a citable sentence. "Premium digital experiences, delivered" is not. The first one answers three fan-out queries at once. The second answers none.
Write the constraints too, including the ones that lose you work. Saying who a service is not for is one of the most quotable things a service page can contain, because it answers a comparison question directly. It also filters your enquiries, which most teams discover they wanted anyway. There is more on the page structure itself in our piece on designing a services page that sells.
Use Organization or LocalBusiness for the business, and describe the offering in the page copy rather than inventing markup for it. Google requires very little. For LocalBusiness the required properties are just name and address, with everything else recommended rather than mandatory.
Google recommends adding telephone, url, priceRange, openingHoursSpecification, and geo coordinates where they apply. It also tells you to "use the most specific LocalBusiness sub-type possible", naming examples like Electrician, Plumber, Store, and Pharmacy. If a sub-type fits your business exactly, use it instead of the generic parent.
The rule that matters most is the one people break. Google says to match your structured data to the visible text. Markup that claims a rating, a price, or a service area the page never mentions is a liability, not an optimization. We have seen sites lose rich results over exactly this, and cleaning it up is usually a one hour job.
Keep it in JSON-LD in the head of the page. In Webflow that means a custom code block on the page or, for templated service pages, a CMS-driven embed so each page emits its own accurate markup rather than one shared copy that is wrong on nine pages out of ten.
With evidence a stranger can check. Named clients, real numbers, dated work, and a visible team. Google is unusually blunt about the priority here. In its guidance on creating helpful content it says of experience, expertise, authoritativeness, and trust, "of these aspects, trust is most important".
That framing is useful because it tells you where to spend effort. A page that says "trusted by industry leaders" with six grey logos proves nothing. A page that says which client, what the situation was, what you did, and what changed proves a great deal. The second version is also far easier for a model to summarise accurately.
Google self-assessment questions push the same way. It asks whether your content "clearly demonstrate first-hand expertise and a depth of knowledge" and whether the content provides "original information, reporting, research, or analysis". A service page can do both. Describe a problem you actually solved and the tradeoff you actually made, and you have cleared the bar most competitor pages do not attempt.
Yes, and the data is not subtle. Pew Research Center analysed the browsing behaviour of 900 US adults across 68,879 Google searches in March 2025. When an AI summary appeared, users clicked a traditional search result on 8% of visits. Without a summary, they clicked on 15%.
The rest of that study is worth sitting with. Pew found users clicked a link inside the AI summary itself on "just 1% of all visits". Sessions ended more often too, on 26% of pages with an AI summary compared with 16% of pages showing only traditional results. Roughly 18% of searches in the study produced a summary at all.
So the honest read is that AI summaries reduce clicks rather than eliminate them, and they are not yet on most queries. The strategic answer is not to give up on clicks. It is to accept that visibility now has two parts, the citation and the click, and to value both. Pew also found that 88% of summaries cited three or more sources, which means being one of several cited pages is a realistic goal rather than a winner-takes-all fight.
For a service business the citation has real value even without the click. Being named as an option in an answer is how a lot of shortlists now get built. The click often comes later, on a branded search.
Write the page around the questions, not around your service names. Buyers ask about price, timeline, process, who else you have done this for, and what happens if it goes wrong. If your page answers those five, it covers most of the fan-out a model will run.
Pricing is where nearly everyone flinches. You do not have to publish a rate card. You do have to say something. A range, a starting point, or an explanation of what moves the number all give a buyer something to work with, and all give a model something to cite. A page with no price signal at all simply drops out of every budget-shaped query.
Timeline works the same way. Saying a build usually takes four to eight weeks answers a question buyers ask before they ask anything else. It also sets expectations that save you a discovery call. In our experience the pages that convert best are the ones that answer the awkward questions before the form, not after it.
Then check whether the page is even reachable. A service page four clicks from the homepage with two internal links pointing at it is competing with one hand tied. We covered how that plays out in our article on site architecture and click depth.
Track impressions and average position in Google Search Console for question-shaped queries, watch referral traffic from AI hosts, and run your own buyer questions through the assistants every month. There is no clean dashboard for AI citation yet, so the measurement is partly manual and that is fine.
Search Console remains the most reliable signal because AI Overviews impressions are folded into web search reporting. A page that gains impressions while clicks stay flat is very often being surfaced in a summary. That pattern is frustrating, but it is information, not failure.
The manual check matters more than people expect. Take the ten questions a good prospect asks on a first call, run them through ChatGPT, Perplexity, and Google AI Mode, and record whether you appear. Do it monthly and you will see movement long before any tool reports it.
Rewrite the first paragraph so it states what you do, who for, and what it costs or takes. That single change does more than schema, more than headings, and more than any AI-specific file. It is also the change most teams resist, because a vague opening feels safer than a specific one.
After that, add the structured data, match it to the visible text, and put two or three real proof points on the page with names and numbers attached. Then leave it alone for a quarter and watch Search Console rather than tinkering weekly.
If you would rather have someone else look at your service pages and say plainly what is missing, that is a conversation we enjoy. Reach out at phoenix.studio and we will go through them with you.
Tell us where you want to go. We'll tell you how we'd get you there.