A real case study on AI search visibility: what we found broken on the client's site, what we rebuilt, and the measurable results that followed.
Most agencies still treat AI search visibility as an SEO add-on, a checkbox tacked onto a ranking strategy once it's already built. We think that's backwards. AI search visibility is now the upstream requirement; traditional rankings follow from it, not the other way around. This case study shows what that looks like in practice: a client with declining rankings and zero AI citation presence, a four-layer rebuild across architecture, content, schema, and checkout UX, and the shift that followed once we treated "getting cited" as the primary goal instead of a side effect.
What Was the Client's Starting Problem?
A regional B2B distributor of outdoor and industrial equipment came to us losing ground on two fronts simultaneously: traditional rankings were sliding for competitive terms, and the site had no presence whatsoever in AI-generated search answers, increasingly where their buyers were starting research instead of Google's classic results page.
The site wasn't broken in any obvious way. It loaded fine, carried a full product catalog, and published blog content on schedule. It was simply built for a search environment that no longer existed.
Three problems compounded each other:
Thin, templated product pages. Titles like "Model X400 – 2HP Motor" told a crawler almost nothing about buyer intent - no use-case framing, no comparison context, no schema depth.
No structured data at all. Zero FAQ schema, no HowTo markup, no product schema. Current estimates put AI Overview coverage of search queries somewhere in the 13%–60% range depending on query type and measurement method, and this site had nothing on the page an answer engine could confidently extract or cite.
A checkout flow quietly capping revenue. Traffic that did convert into product-page visits dropped off at a five-step, desktop-first checkout. This wasn't a visibility problem, it was a "can't cash in the visibility we do have" problem, and it suppressed the ROI of every other fix before we'd even made one.
What Did We Actually Change to Rebuild AI Search Visibility?
We rebuilt the site across four layers at once - information architecture, content, structured data, and checkout UX rather than treating this as a single-lever SEO fix. Fixing only one layer would have left the others capping the result.
Layer | Before | After |
Site structure | Flat category pages, no topical clustering | Hub-and-spoke architecture grouping products by use case, not SKU category |
Content | Spec-sheet product copy | Question-led product and category pages answering real buyer queries (e.g. "what size generator do I need for a 2,000 sq ft workshop") |
Structured data | None | Product, FAQPage, and HowTo schema across the top revenue pages |
Product titles | Model-number-first | Rewritten around search and voice-query phrasing, tested against real query variants |
Checkout/UX | Five-step, desktop-first flow | Two steps, mobile-first, autofill-enabled |
Keyword and Product Title Testing
This is where most AI search visibility projects stop at theory. We didn't. We ran live A/B tests on product title phrasing - model-number-led versus use-case-led against real search and click data, not assumptions. Use-case-led titles consistently outperformed spec-led ones on both organic engagement and AI citation, because they matched how buyers actually phrase questions, not the internal SKU logic the client had defaulted to for a decade.
User Flow and Experience
Visibility gains are wasted if the site loses the visitor once they arrive. Mapping the full path, from AI-cited answer or organic click through to purchase showed the real leak wasn't at the top of the funnel. It was at the checkout. Fixing structure and content without fixing this would have driven better traffic straight into the same broken exit.
What Did the Repositioning Produce?
Organic visibility recovered, the site began appearing in AI-generated answer citations for the first time, and checkout completion improved meaningfully because the UX fix compounded with the traffic-quality gains instead of sitting separate from them.
The traffic volume wasn't the important part. What mattered was that a meaningful share of new sessions came from queries the site had never appeared for before at all, the clearest signal that structured, question-led content does what it's supposed to do: get cited, not just ranked. This is the same shift we've written about in how AI agents are changing the way businesses get found and evaluated online visibility increasingly runs through systems making decisions on a buyer's behalf, not just the buyer scrolling a results page.
FAQ
Does improving AI search visibility hurt traditional Google rankings?
No, the two goals reinforce each other rather than compete. Structured, question-led content that AI systems can cite is largely the same content Google's traditional ranking systems reward, because both are optimizing for clear, authoritative answers to real queries. We haven't seen a case where AI-readiness work traded off against organic rankings; the overlap is close to total.
How long does a repositioning project like this take to show results?
Most measurable movement appears within a few months of implementation. Structural changes - schema, information architecture tend to produce early technical signals fast, while ranking and citation gains build progressively over the following quarters as search engines and AI crawlers re-index and re-evaluate the site.
Do you need to rewrite an entire site to fix AI search visibility?
No, prioritization beats total volume every time. Starting with the highest-traffic or highest-revenue pages and their supporting schema typically produces the fastest, most measurable return, and gives you a proof point before committing to a site-wide rebuild.
What's the actual difference between ranking on Google and being cited by an AI system?
Ranking means your page appears in the traditional results list; citation means an AI system pulls your content directly into its generated answer. A page can rank well and still never get cited if it isn't structured for extraction, which is exactly why schema markup and direct-answer formatting now matter as much as keyword targeting used to.
Is checkout or UX really part of an SEO and AI search visibility project?
Yes, visibility gains are wasted the moment the site can't convert the traffic it earns. Any repositioning project that improves discoverability without auditing the conversion path is solving half the problem and reporting on the easy half.
The Takeaway
Ranking and being the answer stopped being the same job. A site can hold its position in ten blue links and still be functionally invisible in the systems shaping a growing share of buyer research. Treating AI search visibility as the finish line, not SEO with an AI feature bolted on is what separates sites that recover from sites that just keep publishing more content into a structure that was never built to be cited.
Want a breakdown of where your own site is losing AI search visibility? Talk to Abacus Digital.