A resort-wear label went from invisible to cited across ChatGPT, Perplexity and Google Overviews in a quarter. The exact plays we ran.
A mid-size Indian resort-wear brand came to us cited in almost zero AI answers for its core buying questions. Ninety days later it was a named source across ChatGPT, Perplexity, and Google’s AI Overviews for the queries that drive its category, with citation share up more than four-fold. This is the full case study — the starting problem, the exact plays we ran week by week, the results, and how to apply the same system to any D2C brand.
The brand sells premium resort and vacation wear direct to consumers — linen sets, tropical prints, beach-holiday staples — competing in a crowded D2C category where discovery increasingly happens through AI-assisted research. It had solid traditional SEO, a respectable social following, and genuinely good products. On paper it was doing everything right. Yet when its target customers asked AI engines the questions that precede a purchase, the brand was almost entirely absent from the answers, while a handful of competitors were named repeatedly.
This is a pattern we see constantly: a brand that is competent in classic channels but invisible in the answer layer, losing high-intent discovery it does not even know is happening. The customer asking ChatGPT “what are the best resort-wear brands in India?” never sees a rank report; they see an answer, and if you are not in it, you simply do not exist for that customer. The brand’s problem was not quality or effort — it was that its excellent content and products were invisible to the engines making the recommendations.
We began, as always, with an honest audit of the brand’s citation share across its priority prompts. We mapped the real questions buyers ask — “best resort wear brands in India,” “linen sets for a beach holiday,” “where to buy tropical prints online,” and dozens more — and checked, prompt by prompt and engine by engine, whether the brand appeared. The result was stark: cited in roughly one in ten of the tracked answers, and almost never in the highest-value comparative prompts where purchase decisions are shaped.
The audit also revealed who was winning and why. Three competitors owned the answers, and reading their cited passages made the reasons clear: they led with direct, quotable answers, they had clean structured data, and they were corroborated by mentions on publications and community sources the models trusted. The brand had none of these consistently. The audit turned a vague sense of “we should be doing more with AI” into a specific, prioritised diagnosis — and a concrete list of prompts to win.
The diagnosis came down to four gaps, each fixable. First, the brand’s pages buried their answers under brand narrative and styling copy, giving engines nothing clean to lift. Second, its content answered only the obvious head terms, ignoring the fan-out of specific sub-questions the comparative prompts generated. Third, its structured data was thin and inconsistent, so engines could not confidently understand its products or its identity as an entity. Fourth, and most decisively, it had almost no third-party corroboration — its claims lived only on its own site, which the models had no way to verify.
None of these was catastrophic on its own, but together they explained the invisibility completely. The competitors were not doing anything magical; they were simply doing the fundamentals the brand had skipped. That is the encouraging part of most AI-visibility problems: the diagnosis usually points to concrete, achievable fixes rather than some structural disadvantage. We turned the four gaps into a ninety-day plan, sequenced so the fastest wins came first and the compounding work started early.
The first phase was research, because you cannot win prompts you have not mapped. We expanded the initial prompt list into a comprehensive set of the questions buyers actually ask across the funnel — broad discovery (“best resort wear brands”), specific product research (“linen co-ord sets for men”), and comparative and transactional queries (“where to buy tropical print shirts online in India”) — drawing on People Also Ask, autocomplete, the brand’s own customer questions, and the engines’ suggested follow-ups.
We then prioritised ruthlessly by intent and business value, identifying the fifteen prompts most likely to precede a purchase and most winnable given the brand’s authority. This prioritised prompt set became the spine of everything that followed — the target list for content, the denominator for citation-share measurement, and the shared definition of success. Two weeks of disciplined research prevented months of unfocused effort, because every subsequent action was aimed at a specific, high-value prompt rather than at “AI visibility” in the abstract.
With the target prompts defined, we rebuilt the brand’s key pages to be citable. For each priority prompt, we crafted a complete, 40–60 word answer at the top of the relevant page — direct, specific, and quotable — before any brand or styling copy. We restructured the surrounding content into short paragraphs, question-shaped headings, comparison tables for the “best X” queries, and FAQ blocks for the follow-up questions. The goal was simple: for every priority prompt, give the engine a clean passage it could lift and attribute without rephrasing.
We also added the evidence that makes passages citable — specifics about fabrics, fit, and use cases, and honest comparative detail rather than vague superlatives. This was not about writing more; it was about restructuring what existed so its genuine quality became machine-readable and quotable. Within this phase, pages that had buried their value behind narrative became pages that answered the buyer’s question in the first line — the single highest-leverage change in the entire program, and the one that started moving citation share fastest.
Overlapping the content work, we fixed the structural signals. We implemented consistent Product and Review schema across the catalogue so shopping-oriented answers could understand and compare the brand’s items, FAQPage markup on the question blocks, and, crucially, Organization schema with accurate sameAs links tying the brand to its verified profiles and references. This last piece addressed the entity-clarity gap directly, giving the engines an unambiguous definition of who the brand was so they could attribute facts and citations to it confidently.
We validated everything and generated it from the brand’s live product data so it would stay accurate as the catalogue changed. This phase is often underrated because it is invisible to users, but its effect on machine understanding is significant: clean, consistent schema turned the brand from an ambiguous string the engines had to guess about into a clearly-defined entity with clearly-labelled products and answers. Combined with the liftable content, it made the brand not just quotable but confidently attributable — the difference between being pulled into an answer and being named in it.
The final and most decisive phase addressed the corroboration gap. Because models weight the agreement of independent, trusted sources so heavily, we ran targeted digital PR to seed accurate mentions of the brand on the publications and community sources its customers — and the engines — already trusted. This was not link-spam; it was earning genuine, accurate coverage and presence that turned the brand’s claims from unverifiable self-assertion into facts the models could corroborate and repeat with confidence.
This phase compounds slowly and pays the largest dividends, which is why we started it early and let it run through the end of the program. As accurate third-party mentions accumulated, the brand crossed a threshold: the engines began treating it as an established, corroborated option rather than an unknown, and its citation share on the competitive comparative prompts — the ones that had been hardest to crack — started to climb. Corroboration was the factor that moved the brand from “sometimes cited on easy prompts” to “named on the prompts that drive purchases.”
Throughout, we measured citation share continuously across the full prompt set and all target engines, so we could see movement in real time and reallocate effort toward what was working. We tracked the brand’s share against the three incumbent competitors, watched branded search and direct traffic as leading indicators of created demand, and monitored the small but telling stream of AI-referred sessions. This measurement was not an afterthought; it was the control system that let us steer the program week by week rather than hoping at the end.
The continuous view mattered because it caught both wins and gaps early. When a restructured page started getting cited, we saw it within days and applied the same pattern to similar pages; when a prompt stubbornly resisted, we could see that corroboration, not content, was the missing factor. Measuring citation share as a live signal — the exact capability DUNkē provides — turned the program from a set of hopeful actions into a managed, evidence-driven process.
Citation share across the fifteen priority prompts rose from roughly 9% to 38% — more than a four-fold increase — with the brand moving from absent to named on the majority of its highest-value comparative queries.
Branded search climbed around 60% over the same period as more customers encountered the brand in AI answers and later sought it out. AI-referred sessions, though modest in volume, converted at nearly double the site average — unsurprising, since those visitors arrived already informed and predisposed by the answer that named the brand.
Just as importantly, the gains held and kept compounding after the ninety days, because the corroboration and entity signals we built are durable assets, not rented placements. The brand had moved from invisible to established in the answer layer of its category — a position that is genuinely difficult for competitors to dislodge, precisely because it rests on accumulated trust rather than a temporary tactic.
Reflecting on the program, the results came from nothing exotic — just the fundamentals, run as a coherent system and measured properly. The answer-first restructuring produced the fastest early wins by making the brand’s existing quality quotable. The schema and entity work made the brand confidently attributable. And the corroboration, though slowest, was the factor that unlocked the competitive prompts and made the gains durable. No single play was sufficient alone; the compounding came from doing all of them, in sequence, against a prioritised prompt set.
The other decisive factor was measurement. Tracking citation share continuously meant we were never guessing — we could see which plays moved which prompts and double down accordingly, and we could prove the program’s value with a number that mattered. A brand running the same tactics without measurement would have flown blind and struggled to justify the investment; measurement turned the effort into a managed, demonstrably successful program.
In the spirit of honesty, not every part of the program was optimal. We would start the corroboration work even earlier, since it is the slowest-compounding and highest-impact lever, and beginning it in week one rather than week seven would likely have accelerated the competitive-prompt gains. We would also invest more upfront in the brand’s own original content assets — a styling guide, original data on resort-wear trends — because those become the corroboration magnets that earn mentions naturally over time.
These are refinements, not regrets; the program succeeded. But they reflect a general lesson: the durable, compounding parts of AI-visibility work — corroboration and genuine authority-building — should start as early as possible, because they take the longest to pay off and pay the most in the end. The fast wins from restructuring are satisfying, but the lasting position comes from the slow work, and starting it sooner is almost always right.
The brand’s story generalises directly, because its starting problem is the default state for most D2C brands right now: competent in traditional channels, invisible in the answer layer, and unaware of the high-intent discovery being lost. The path out is the same system we ran — map the prompts that precede purchase, rebuild pages to be liftable, add clean schema and entity clarity, earn corroboration on trusted sources, and measure citation share continuously so you can steer and prove it.
None of it requires a bigger budget than a serious brand already spends on marketing; it requires redirecting effort toward the channel where discovery is moving. The brands that do this now, while the space is uncrowded, build a compounding position that later entrants will struggle to overtake — just as this one did. The opportunity is not that the tactics are secret; it is that most competitors have not yet acted on them.
sameAs.To make the system concrete, it helps to zoom in on a single prompt. Take “best resort wear brands in India” — a high-value comparative query the brand had been entirely absent from at the start. Winning it required all four levers working together: a genuinely useful, answer-first passage that named and briefly characterised the strongest options honestly (including the brand, on its real merits), clean structure the engines could lift, Organization and Product schema so the brand was a clearly-understood entity, and — decisively — corroborating mentions on publications and community threads that discussed the category.
What made this prompt winnable was refusing to treat it as a place for a thin self-promotional page. The passage that eventually got cited read like a fair, informed answer to the question, which is exactly what the engines reward and what a savvy buyer trusts. The brand earned its place in the answer by helping to answer the question well, not by shouting loudest — and once the corroboration caught up, the engines grew comfortable naming it alongside the incumbents. That single prompt, once won, became a steady source of high-intent discovery.
A case study like this can make it sound as though visibility is purely a marketing-mechanics exercise, but the brand’s genuinely good products were a quiet precondition for the whole effort. The comparative prompts that matter most — “best X,” “which brand for Y” — are, in effect, quality judgments, and the corroboration that unlocks them tends to accumulate around brands that are actually worth recommending. Good products earn the organic mentions, the positive community discussion, and the honest inclusion in “best of” content that feed the models’ trust.
This matters because it sets a realistic expectation: AI-visibility work amplifies genuine quality rather than manufacturing it from nothing. A brand with weak products would have struggled to earn the corroboration that carried this one across the competitive prompts, because the sources the models trust would have had little reason to recommend it. The lesson is not that products alone win — the brand had good products and was still invisible — but that the system works best, and the gains prove durable, when there is real substance underneath for the visibility to reflect.
The easy wins came quickly — restructuring pages got the brand cited on straightforward informational prompts within weeks. The hard part, and the real test of the program, was the competitive comparative prompts the incumbents owned, where being merely present was not enough and corroboration was the deciding factor. These prompts resisted content changes alone; the brand could publish the best passage in the world and still be omitted if the models had no independent basis to trust it as a genuine option in the category.
Breaking through required patience and the corroboration work compounding over the full ninety days. As accurate third-party mentions accumulated — coverage, community discussion, honest inclusion in category round-ups — the brand crossed the threshold from unknown to established, and the engines began naming it alongside the incumbents on exactly the prompts that had been hardest to crack. This was the most satisfying part of the program precisely because it was the hardest, and it is where the durable competitive advantage was built — a position resting on accumulated trust that competitors cannot quickly erase.
It is worth being concrete about the effort, because “four-fold in ninety days” can sound like magic when it was in fact disciplined work at a sensible scale. The program ran over three months with a focused team: prompt research and strategy up front, a content sprint to rebuild the priority pages, a technical workstream for schema and entity clarity, and an ongoing digital-PR effort for corroboration — all coordinated against the prioritised prompt set and steered by continuous citation-share measurement. It was not a massive budget; it was existing marketing effort redirected toward the channel where discovery was moving.
The sequencing was as important as the scale. The fast wins from restructuring built early momentum and stakeholder confidence, while the slow-compounding corroboration work — started as early as we could — matured in time to unlock the competitive prompts by the end. A brand attempting the same thing should expect a similar shape: quick initial gains from content and structure, followed by a steadier climb as corroboration and entity signals accumulate. Ninety days is enough to prove the system on a focused prompt set, which is exactly what this program did.
The numbers describe the outcome, but the changed customer journey is what they mean in practice. Before the program, a customer researching resort wear through AI simply never encountered the brand — it was absent from the answers that shaped their shortlist, and the discovery was lost silently. After, that same customer asking the same questions saw the brand named as a credible option, often alongside or ahead of the incumbents, with the engine’s implicit endorsement. The brand had inserted itself into the consideration set at the exact moment it was being formed.
That earlier, endorsed presence is why the downstream effects followed — the branded search as customers remembered the name and looked it up, the direct visits, and the higher conversion of visitors who arrived already predisposed by the answer that had recommended the brand. The citation did not just add a touchpoint; it added a trusted one at the top of the journey, which is worth far more than an untrusted impression later. The whole point of the program was to be in the room when the decision was being shaped, and the changed journey is the evidence that it worked.
It is achievable but not guaranteed — the starting point matters enormously. A brand beginning from near-invisibility, as this one was, has the most room to grow, so large multiples are realistic. A brand already moderately visible will see smaller multiples on a higher base. The right expectation is meaningful, durable improvement on a focused prompt set within a quarter, not a fixed number.
No — the system is category-agnostic. The specific prompts and corroboration sources differ, but the pattern — map high-intent prompts, build liftable answers, add schema and entity clarity, earn corroboration, measure continuously — applies to any brand whose buyers research through AI, from SaaS to services to other retail categories.
The answer-first restructuring produced the fastest wins, but corroboration was the decisive factor for the competitive, purchase-driving prompts — and it compounds slowest, so starting it early matters most. No single lever was sufficient alone; the results came from running all of them as a coordinated system and measuring the whole way.
“You don’t win the answer layer with a trick. You win it with a system — the right questions, answered better than anyone, on pages machines can read, echoed by sources they trust — and measured every step of the way.” The Age’X
Get a free GEO audit — the same analysis behind every article here.