# They don't search for you. They ask.

> An interactive Dashcrab story about AI visibility (GEO). Text version of https://dashcrab.app/en/stories/te-preguntan.html
> Published October 12, 2026. More stories: https://dashcrab.app/en/stories.html

Ñandú Outdoor ranks first on Google for its keyword. When someone asks an AI which jacket to buy, it doesn't show up. Why the game moved from keywords to topics.

## Before: one keyword, one page, one position

Mateo does SEO for Ñandú Outdoor. His pride: first on Google for "women's waterproof jacket". That's how the game was played for twenty years.

## Now: a question with context

Sofi doesn't search like that. She asks ChatGPT: "I'm going to El Chaltén in April, I get soaked easily and I have USD 150. Which jacket should I get?".

The AI doesn't run one search with that question. It breaks it into subtopics and searches each one separately. Example searches and who answers them:

| Search | Who answers it |
|---|---|
| El Chaltén weather April | a weather site |
| waterproof or windbreaker jacket | a trekking guide |
| best waterproof jackets 2026 | a magazine |
| what is 10,000 mm water column | a mountain forum |
| women's waterproof jacket | **nandu-outdoor.com** |
| waterproof jackets under 150 dollars | a price comparison site |
| Cumbre Sur Ventisca reviews | Reddit |
| layering for trekking in Patagonia | YouTube |

Ñandú answers just one. The answer recommends three jackets, two from Cumbre Sur and one from Viento Austral, citing a weather site, a magazine, Reddit and YouTube. None of them is Ñandú's.

### How it works: query fan-out

Google documents that AI Overviews and AI Mode may use a "query fan-out" technique: they run several related searches, across different subtopics and sources, to build a single answer. Google hasn't published how many searches it runs per question. ChatGPT also rewrites and searches several times; the GEO Observer extension shows those internal searches.

### Study: SEO gets you halfway

Across 518 real Google searches (July and August 2026), 47.45% of what the AI Overview cited was already in that search's organic results. In AI Mode, 85.29% of citations were in neither the organic results, the ads nor the AI Overview: only 1 in 5 organic results survives (Dashcrab, AI citation study 2026).

### Fact: 96.7% is someone else talking

On average, a brand's own site gets 3.3% of the citations in its own category, and none went above 10%. The other 96.7% goes to third parties: media, comparison sites, forums, reviews and videos (Dashcrab, AI citation study 2026, 21,571 citations across six models).

### Behavior: they don't even click

Pew Research tracked 68,879 searches by 900 adults. With an AI summary on screen, clicks on a traditional result dropped from 15% to 8%, and the sources the summary cited got a click in 1% of visits.

## From keywords to topics

Mateo thought in keywords. The AI thinks in topics: everything you need to know to answer well. For "trekking jacket for Patagonia", the subtopics are: weather by month and area; layers (base, warmth, rain); water column and membranes; comparisons by price; sizes and fit; care and washing; forum opinions and reviews; shipping and returns.

Ñandú covers 2 of 8 (sizes and shipping). Cumbre Sur covers 6. That's why the AI cites it.

### Study: what moves the needle

The paper that named GEO (Aggarwal et al., KDD 2024) tested content changes on about 10,000 questions. Adding statistics, quotations and sources raised visibility in the answers by up to 40%; keyword stuffing didn't help. The effect varied by industry: there's no single recipe.

### Concept: SEO isn't going away, something gets added

Almost half of what the AI Overview cites already ranked: SEO still opens the door. But the AI cites inner pages, not the homepage: between 67.8% and 75.2% of citations go to pages two or more levels deep. The new strategy is covering the whole topic, with a page that answers each subtopic, and being where the AI goes to look: reviews, forums and comparisons.

## Measured, engine by engine (example data)

Ñandú's position in each engine's answer.

| Question | ChatGPT | Claude | Gemini | Perplexity | AI Mode |
|---|---|---|---|---|---|
| Which jacket should I get for El Chaltén in April? | no (Cumbre Sur appears) | no (Cumbre Sur) | no (Cumbre Sur) | no (Cumbre Sur) | no (Cumbre Sur) |
| Best waterproof jacket under 150 dollars | #3 | no | #4 | #2 | no |
| Is Ñandú Outdoor a good brand? | #1 | #1 | #1 | #1 | #1 |
| Women's waterproof jacket | no | #5 | #2 | no | #1 |
| How to choose a jacket's water column | no | no | no | no | no |

Dashcrab tells you why each question was lost in each engine: whether it didn't read you, read you and didn't cite you, or the conversation runs through reviews and forums where you aren't.

### Careful: each engine cites something else

Comparing citations across five models, they agree on 3.5% of exact URLs and 10.4% of domains. ChatGPT cites more product pages; Claude, Gemini, Perplexity and AI Mode, more blog posts. There's no page that wins everywhere.

## Three things to do this week

1. **Write 20 questions, not keywords.** The way a customer would ask them, with context. Ask them in ChatGPT, Gemini and AI Mode and note who shows up.
2. **Map the subtopics of your main topic.** What the AI will search for to answer. Do you have a page that answers each one, with data and sources?
3. **See who the AI cites in your industry.** Reviews, forums, comparisons, videos. You need to be there too.

SEO puts you on the list. The topic puts you in the answer. Every week Dashcrab asks what your customers ask in ChatGPT, Claude, Gemini, Perplexity and Google AI Mode: https://dashcrab.app/en/geo.html. The GEO Observer extension is free: https://dashcrab.app/en/extension.html

## Sources

- Google Search Central, AI features and your website: https://developers.google.com/search/docs/appearance/ai-features
- Dashcrab, AI citation study 2026: https://dashcrab.app/en/research/ai-citation-study-2026.html
- Pew Research Center (2025): https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/
- Aggarwal, Murahari, Rajpurohit, Kalyan, Narasimhan and Deshpande, "GEO: Generative Engine Optimization", KDD (2024): https://arxiv.org/abs/2311.09735