The challenge
The brand was profitable, well-reviewed on its own site, and ranked respectably on Google. But buyers in the 25-to-34 segment had quietly moved their product research into ChatGPT and Perplexity — and the brand was almost never cited there. A competitor with weaker product but stronger citation surface was getting recommended on roughly 6 in 10 generative queries in the category.
The founder had spent the previous quarter doubling down on traditional SEO, with diminishing returns. The diagnosis was wrong — the search demand had not disappeared; it had migrated to surfaces the SEO program could not reach.
The approach
01
Citation surface audit
We ran 240 buyer-intent queries against ChatGPT, Gemini, Perplexity, and Claude — across product, comparison, and "best of" categories. We mapped every source the models cited and ranked them by retrieval weight. Three Reddit communities, two niche review blogs, and one comparison site accounted for 71% of the brand mentions in the category.
02
Answer-engine asset production
We produced 18 assets engineered for LLM retrieval — answer-first paragraphs, structured comparison tables, schema-marked FAQs. The angle was not "best skincare brand" (a saturated query) but the long-tail, problem-led queries that buyers actually run through LLMs ("what helps with hormonal breakouts in your thirties").
03
Authority distribution
We placed the assets on the high-weight sources our audit had surfaced — not the high-DA-but-low-LLM-weight sites a traditional SEO program would target. Three of the placements went live within two weeks; the others rolled out over the quarter.
04
Founder voice on social
Parallel to the GEO work, we built a LinkedIn and short-form video presence for the founder around the same long-tail problem-led content. The social work seeded the conversations the LLMs would later retrieve from — a slow-build flywheel that compounded with the direct GEO work.
05
Weekly LLM ranking measurement
We ran the same 240 queries weekly and tracked where the brand appeared, in what context, and against which competitors. The data fed back into the next round of asset production.
The outcome
In 90 days the brand mention rate inside ChatGPT and Perplexity rose by 47%, and share-of-voice against the lead competitor inside Perplexity reached 4.1x. Organic-attributed revenue — measured through last-touch attribution and post-purchase surveys — grew 31% over the same window.
Qualified lead cost fell by 52%, driven primarily by the founder-led social engine becoming a top-of-funnel source. Most of those leads cited "I saw this in ChatGPT" or a founder post as the reason they bought, in the post-purchase survey.
Lessons
- A weak citation surface is invisible to traditional SEO tools but devastating in the LLM era. Audit it explicitly.
- Long-tail, problem-led content outperforms broad category content inside LLMs — the model rewards specificity.
- GEO and organic social compound. The social content seeds the conversations LLMs later retrieve from.
We thought we had an SEO problem. We had a citation problem. Ninety days in, the change in how often we get mentioned is the single biggest unlock our marketing function has had this year.
Founder, D2C Skincare Brand