Most GEO case studies start with a brand that's already ranking somewhere and just needs to be cited more often. Zenity's starting point was blanker than that: agentic AI security, the category it sells into, barely existed as something buyers searched for. There was almost no keyword volume to target and zero AI citations to build on. Six months later, four AI engines were naming Zenity as the answer. This is the content walkthrough behind that jump.
The category that didn't exist yet
Agentic AI security is the kind of category that makes classic keyword research nearly useless. There wasn't a stable set of search terms to target because buyers weren't yet searching in a consistent, established vocabulary, they were asking AI engines directly to explain a category that was still forming in real time. When they did, the generated answers named analysts and larger, more established competitors. Zenity wasn't in the conversation at all.
The technical foundation compounded the problem. The site carried thin schema, a slow crawl budget, and no machine-readable structure built for AI crawlers specifically. Even the strong material Zenity had already published couldn't get parsed cleanly enough to be quoted, so genuinely good content was sitting invisible in exactly the place buyers were looking first.
Zero mentions in AI answers and close to zero measurable monthly search volume for the category is a genuinely uncomfortable place to start an engagement, because it removes most of the usual diagnostic tools a team would normally lean on. There's no keyword-gap report to run when the keywords barely exist yet, and no competitor content audit that tells you much when the competitors named in AI answers were analyst firms and platform incumbents rather than direct category peers. The only real signal available at the start was qualitative: what were buyers actually asking, in their own words, when they typed a question into an AI engine instead of a search box, and what did the engine say back.
That qualitative starting point is what shaped the entire six-month sequence. Instead of optimizing for a keyword list, the work optimized for a set of real buyer questions, logged directly from probing the engines the way a prospective customer would, and then built content specifically to answer those questions better and more citably than whatever the engines were currently pulling from.
Month by month: how the citation curve moved
Rather than chase a handful of low-volume keywords, the approach treated the entire category as the opportunity: if buyers were going to ask an AI engine to define agentic AI security, the plan was for Zenity's content to be the definition the engine reached for.
The overlap between phases two and three is deliberate and worth calling out, because it's the part teams most often get backwards. Category content shipped before the GEO layer was fully in place, not after, on the logic that there was nothing for llms.txt and answer blocks to point at until the definitional content existed. Machine access without content to access isn't a foundation. It's an empty room with a very clean door.
The month-five-and-six "compound and tune" phase is easy to undersell in a summary, but it's where the earlier work actually converted into the headline numbers. Daily citation monitoring across four engines meant the team could see, within days rather than months, which specific pages an engine had started quoting and which ones it kept skipping. Pages that were getting picked up got expanded and interlinked more aggressively. Pages that weren't landing got rewritten with a sharper answer block, sometimes more than once, until the pattern that was working elsewhere on the site got applied there too. That iterative loop, ship, measure per-engine, adjust, is the part of GEO work that a single before-and-after case study number tends to hide, and it's also the part that's hardest to shortcut.
What actually changed technically
The technical rebuild wasn't cosmetic. A GEO-readiness score built for the engagement moved from 31 to 94 over the six months, tracked against four specific, previously-failing checks.
| WHAT WAS FAILING | THE GAP | THE FIX |
|---|---|---|
| Answer extractability | Key definitions were buried in prose engines had to paraphrase to use | Restructured into lead answer blocks engines could quote verbatim |
| Schema coverage | No Article or entity schema, so engines couldn't reliably resolve the page | Full Article, FAQ, and entity markup deployed across templates |
| llms.txt | No machine-readable map for AI crawlers to follow | Shipped and kept in sync with the sitemap |
| Crawl budget | Slow rendering wasted crawl budget on low-value pages | Fixed render performance and pruned low-value crawl paths |
Fixing "answer extractability" is the row that mattered most for the citation curve specifically. A page can carry the correct information and still never get quoted if that information is buried three paragraphs into dense prose an engine has to paraphrase rather than lift directly. Restructuring the same underlying facts into a lead answer block, without changing what was actually being said, was frequently the difference between a page an engine ignored and the same page an engine started quoting within weeks.
None of these four fixes were exotic. Schema, crawl budget, and llms.txt are the same basics we outline in llms.txt, explained: a plain-text map that tells an AI crawler who a business is, what it does, and where its most citable content lives. What made the difference for Zenity wasn't a novel technique, it was doing all four at once, in the right order, against a site that had never had any of them addressed, which is a large enough combined gap that fixing it produces a step change rather than an incremental one.
The content architecture behind it
The content strategy leaned on the same hub-and-spoke model that works for any B2B category with a multi-stakeholder buying committee, adapted for a category still being defined in real time. The hub held the category definition itself, effectively becoming the page an AI engine could point to when a buyer asked "what is agentic AI security," while the spokes handled the more specific, comparison-heavy questions a security buyer actually researches once they understand the category exists.
Those spokes leaned heavily on comparison and alternatives content, which lines up with what we've found holds true across categories more broadly: comparison content earns a disproportionate share of AI citations because it answers the buyer's first real question, what are my options, in a format an engine can lift directly. In a brand-new category, that comparison content was doing double duty: it was simultaneously teaching the market what the category even contained and positioning Zenity as a named option the moment that understanding formed.
Once the GEO layer shipped in month three, the citation curve moved from single digits to consistent quoting fast, engines increasingly treating Zenity's hub as the category's reference source rather than one voice among several. That's the compounding effect a category-definition strategy is built to produce: the earlier a brand becomes the source an engine already trusts for the basics, the more of the follow-up, higher-intent questions default back to the same source.
It's worth being specific about why comparison content carried so much weight in a category this new, because the reasoning is different from applying it to an established market. In a mature category, comparison pages mostly answer "which of these known options should I pick." In a category still forming, the comparison pages were also doing the work of establishing which options exist at all, effectively defining the competitive set a buyer should even be considering. Getting cited as part of that competitive set, rather than left out of it entirely, was worth more in month two of this engagement than a dozen additional blog posts optimized for search terms with a still-forming, unstable search intent behind them.
The client's own read on this, delivered once the results were in, is a fair summary of what changed: they didn't chase keywords, they made Zenity the definition of the category. That's a different kind of win than a typical SEO engagement produces, because it's not really competing for position within an existing, well-understood market. It's about being early enough and structured well enough that the market's own reference points get built around you instead of around whoever else was already there.
What this means if your category has no citations yet
Most brands aren't starting from Zenity's exact position, a category with almost zero search volume and zero prior citations, but the sequencing holds even for an established category with weak AI visibility: fix machine access first, ship category-defining content before the GEO layer, then let daily citation monitoring tell you which pages are actually landing so month five and six can double down on what's working instead of guessing.
The score that moved from 31 to 94 is also worth treating as a template rather than a one-off metric. Any team can run the same four-check version of that audit against its own site today, answer extractability, schema coverage, llms.txt presence, and crawl budget efficiency, and get an honest read on how much of this six-month curve is available to them without waiting for a category to form first. Most established brands will find at least one of those four checks failing outright somewhere, and unlike building an entirely new category from nothing, fixing an existing gap in an already-searched category tends to move meaningfully faster, not slower, once the underlying technical work is actually done.
For cybersecurity and other technical categories where buyers increasingly ask an AI engine to explain the category before they ever read a vendor's homepage, being the source that gets cited is worth more than ranking below the fold for a handful of established keywords. If your category is being defined in real time and nobody's claimed the definition yet, that's the opportunity this walkthrough is describing, and it's exactly the kind of engagement our content marketing and GEO teams run together rather than in sequence, because in a new category, the two aren't separable.
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Tyler leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.