All insights
XINDAR INSIGHT

How the Giants Play: Seven Case Studies in Real-World AI Search Strategy

The New York Times earned 1,890 LLM citations. A sauna studio 89. Seven documented cases reveal what actually moves AI search visibility — at every budget.

Strategy pieces have a standard weakness: they describe mechanisms without showing receipts. This piece collects seven documented cases — from a century-old newspaper to a nine-month-old local business — where AI-search-era tactics produced measured results. The cases are not equally rigorous (some are self-reported; the caveats section says so plainly), but together they form the closest thing the field currently has to an evidence base.

1. The New York Times: The Citation Colossus

Detailed.com's analysis of how major publishers appear inside LLM answers found the New York Times had accumulated roughly 1,890 citations across large-model answer sets — the largest footprint in its dataset. The mechanism is not mysterious: a century of accumulated authority, an enormous daily publishing cadence, and content that gets referenced across the web makes the Times the answer to a vast range of queries by default.

The lesson for everyone else is negative space: you cannot out-publish the Times, but you can study what the citation flow rewards — original reporting, named expertise, and being referenced by other publications. News organizations alone account for roughly 27% of AI citations (per the Muck Rack analysis covered earlier in this series). The giants' citation share is the visible tip of an earned-media logic that applies at every scale.

2. Forbes: The Refresh Production Line

Forbes increased content-refresh output by 41% as a deliberate AI-era operating decision — treating the update of proven URLs as a production system rather than occasional maintenance. In a search layer where (per the 5.3M-result study in our freshness piece) freshness is the one attribute predicting top-3 visibility, the largest publishers are re-manufacturing their archive. The strategic read: refresh capacity is now a competitive moat, and the giants are already staffing for it.

3. Business Insider: Share of Voice as a Business Model

Business Insider maintained roughly 27% share of voice in its monitored answer categories — a measure of how much of the conversation about its beats flows through its coverage. The instructive part is the posture: an organization built around being quoted rather than merely read. In an answer layer where 84% of citations flow to earned media, publishers that engineer quotability — distinct data, clear attributions, defensible scoops — convert coverage into citation share at scale.

4. The Financial Times: Conversion, Not Just Visibility

The FT reported conversion rates up 290% on AI-referred audiences in its subscription funnels — a number that reframes the "AI traffic is tiny" debate. Volume is small; intent quality is not. This matches the pattern documented across the field (LLM-referred visitors converting at 2–4× organic rates in multiple verticals). The lesson: judge AI-search channels by conversion-weighted economics, not session counts — the giants already do.

5. Duolingo: AI-Native Production at 10x

Duolingo scaled content output roughly 10x with AI-assisted production workflows (Fortune documented hundreds of AI-assisted stories in a single quarter). The honest reading is double-edged: production scale is now cheap for everyone, which means output volume is no longer a differentiator — the differentiators are what AI cannot manufacture: proprietary data, first-hand testing, genuine expertise. Duolingo's case matters less as "publish more" and more as "the floor has risen; the ceiling now requires things machines can't fake." This is the exact boundary Google's quality framework draws — the tool is fine; the absence of value is not.

6. iPullRank: The Enterprise Case for Content Engineering

Mike King's agency documented engagement-scale results from retrieval-engineered content strategies: 2.4 billion in incremental revenue** for a financial-services client, **290 million for an e-commerce program, and a 130% traffic recovery for an automotive publisher hit by algorithmic suppression. The methodology — engineering content for how retrieval actually selects and extracts passages, rather than for blue-link rankings — is the enterprise-grade expression of the mechanics this series has described. The caveat is the field's standard one: agency case studies are self-selected, and attribution at this scale involves judgment. But the pattern — measurable commercial outcomes from passage-level engineering — is consistent across independent teams.

7. The Sauna Studio: Proof at the Small End

The most useful case for most readers is the smallest. A UK sauna installation studio, working with the agency The 66th, documented roughly 120× traffic growth and 89 AI citations over nine months — in a niche where no incumbent authority existed. Separately, Goodie's client roster reports Dermalogica AI-search visibility up 85%, a UK wealth manager (Rathbones) up 106%, and SteelSeries achieving 3.2× AI-channel conversion over six months.

These small-brand cases carry the series' most consistent structural finding into practice: in the answer layer, the field is frequently empty. Local and niche categories have no incumbent citation ecosystem — the comparison content, review presence, and technical documentation that AI answers need simply does not exist yet. The first credible entrant becomes the default source. That window is why a sauna studio can post triple-digit growth while a mid-market brand with strong classic rankings posts none.

The Pattern Across Seven Cases

Strip away the org sizes and five properties recur:

  1. Earned-media logic dominates. The giants win citations through being referenced, not through on-page tricks — consistent with every citation-pool study in this series.

  2. Refresh is infrastructure. Forbes's 41% is a staffing decision, not a tip.

  3. Conversion-weighted economics favor AI channels. The FT's 290% and SteelSeries's 3.2× say the same thing at different scales: small volume, high intent.

  4. Production scale is commoditized; distinctiveness is not. Duolingo's 10x raises the floor — the ceiling still requires proprietary data and first-hand expertise.

  5. Empty niches are the arbitrage. The sauna case generalizes: where no citation ecosystem exists, a modest, disciplined build becomes the default answer.

The Honest Caveats

Stated plainly, because this piece leans on commercial evidence:

  • Agency and vendor case studies are self-reported and self-selected. iPullRank, Goodie, and The 66th all publish their own results. The direction of the findings is corroborated by independent research elsewhere in this series; the specific magnitudes deserve discount.

  • Publisher figures mix official statements and third-party estimates. The FT and Business Insider numbers come from Detailed's analysis and corporate communications; methodology varies.

  • Correlation discipline applies throughout. None of these cases isolates causation experimentally. The strongest causal evidence in the field remains the 2026 randomized AI Overview experiment covered in our click-economics piece — none of these cases reaches that bar.

  • Survivorship is real. For every documented success there are undocumented failures; no one publishes those.

The Bottom Line

The evidence base for AI search strategy is maturing from theory into cases — and the cases, from the Times to the sauna studio, converge on a consistent picture: earned-media logic, maintained content, conversion-weighted measurement, and the arbitrage of empty niches. The giants are executing these plays with production-line discipline; the small brands are executing them into categories where nobody has bothered.

The gap between those two groups — the mid-market with strong classic rankings and no answer-layer presence — is where most readers of this series operate. The receipts above say the same thing the mechanisms said: the window in your category is probably still open. It will not stay open, because the sauna studios of the world are already walking through it.

Sources

  • Detailed.com (Glen Allsopp), publisher and brand AI-operations analysis (2026): NYT ~1,890 LLM citations; Business Insider ~27% share of voice; FT conversion +290%; Duolingo ~10x output (Fortune: 600+ AI-assisted stories in a quarter); Forbes refresh +41%; "How 16 Companies Dominate Google" archive (500+ properties controlled by 16 firms); Rightmove Q2 2026 earnings (85%+ direct + organic traffic)

  • iPullRank (Mike King) case disclosures:

.4B financial services, 90M e-commerce, 130% algorithmic recovery

  • The 66th case documentation: sauna studio ~120× traffic, 89 AI citations over 9 months

  • Goodie AI client cases: Dermalogica +85% AI search visibility; Rathbones +106% AI citations; SteelSeries 3.2× AI-channel conversion over 6 months; NoGood +335% AI traffic

  • Cross-context series data: Muck Rack 84% earned media / 27% news citations; Growth Memo 5.3M-result freshness study; LLM referral conversion multipliers (Amsive, WebFX); 2026 randomized AI Overview click experiment

  • Case-study figures are self-reported by the organizations named or drawn from third-party analysis of public behavior; treat magnitudes as directional. All data as of 2026.

    Back to insightsMarkdown version