How the Query Fan-Out Explorer Works
These Are Estimates
No AI search engine publishes its real fan-out. Google says AI Overviews and AI Mode may use a query fan-out technique, issuing several related searches across subtopics. We combine the closest public signals with model estimates and label every row with where it came from, so you can see how much to trust it. Rows marked Observed are searches a Google model actually ran through the Gemini API, which is related to, but not the same as, AI Mode.
Sources
- Two model families (claude-sonnet-5-5 and gpt-6.1-terra) each propose 25 to 40 sub-queries across 13 facets. Agreement between them raises stability.
- Gemini with Google Search (gemini-3.8-flash) answers the topic; the searches it actually ran are recorded as Observed.
- Google AI Mode (through DataForSEO): the answer, its sources and the refinements it offers, for the topic and its top follow-up.
- Google results (DataForSEO): People Also Ask (opened several levels deep), related searches, AI Overview sources, autocomplete for six prompts, and the results for the top ten sub-queries.
- Perplexity: the searches it ran to answer the topic, the sources it found, and Reddit threads found through Perplexity Search (we keep only the link, title and a snippet for 48 hours and never call Reddit directly).
Evidence Labels
| Observed | Gemini ran it through Google Search, Perplexity ran it, or Google AI Mode offered it. |
| Demand | People Also Ask, autocomplete, related searches or Reddit show real people ask it. |
| Cited | Perplexity cites sources for it. |
| Model estimate | Only a language model proposed it. |
Candidates are merged when they mean the same thing (embeddings, cosine similarity 0.88 or more), then a fast model assigns the final facet, removes off-topic rows and catches duplicates the embeddings missed.
Stability and Priority
Stability (0 to 100) adds points for each independent source family that produced the sub-query:
| Claude proposed it | 10 |
| OpenAI proposed it | 10 |
| Gemini ran it through Google Search | 25 |
| Google AI Mode offered it | 20 |
| Perplexity ran it | 15 |
| People Also Ask | 15 |
| Autocomplete or related searches | 10 |
| Perplexity cites sources for it | 5 |
| Reddit discussions match it | 5 |
Priority is stability times a facet weight: 1.2 for comparisons, pricing and alternatives, 1.0 for how-to and use cases, 0.8 for the rest. The top 40 rows are shown first.
Coverage Score
When you add a page, we fetch it from our server (public addresses only), extract the main content and split it into passages by heading. For each sub-query we take the three passages closest in meaning and a model judges it covered, thin or missing, suggesting a section to add. The score is (covered + half of thin) divided by the total, weighted by priority. Page text is treated as data, never as instructions to the model. Coverage measures whether the page answers these sub-queries. It does not guarantee AI citations.
Limits and Privacy
Signed-in users get 5 runs a day, resetting at midnight UTC. A repeat of the same topic, market and language within seven days reuses the earlier result and still counts as a run. If every source fails, the run is not counted. We store your email, your runs and what each run cost us; results are private unless you share them, and your page address is never shown on a shared result.
Prices Behind Every Run
Each call is logged with its cost, from these list prices (dated, so past runs stay correct when a price changes):
| Item | From | Price (USD) |
|---|---|---|
| anthropic:claude-haiku-5-5 | 2026-01-01 | $0.1 in / $0.5 out per 1M tokens |
| anthropic:claude-sonnet-5-5 | 2026-01-01 | $2 in / $10 out per 1M tokens |
| openai:gpt-6.1-terra | 2026-01-01 | $1 in / $5 out per 1M tokens |
| openai:gpt-5.6-luna | 2026-01-01 | $0.2 in / $1.2 out per 1M tokens |
| openai:text-embedding-3-small | 2026-01-01 | $0.02 in / $0 out per 1M tokens |
| google:gemini-3.8-flash | 2026-01-01 | $0.75 in / $3.75 out per 1M tokens |
| google:gemini-3.8-flash | 2027-01-01 | $1.5 in / $7.5 out per 1M tokens |
| google:search-grounding | 2026-01-01 | $0.014 each |
| perplexity:sonar | 2026-01-01 | $0.25 in / $2.5 out per 1M tokens |
| perplexity:web_search | 2026-01-01 | $0.0025 each |
| perplexity:search | 2026-01-01 | $0.005 each |
| dataforseo:organic-live | 2026-01-01 | $0.002 each |
| dataforseo:paa-click | 2026-01-01 | $0.00015 each |
| dataforseo:ai-overview-async | 2026-01-01 | $0.002 each |
| dataforseo:autocomplete-live | 2026-01-01 | $0.002 each |
| dataforseo:ai-mode-live | 2026-01-01 | $0.004 each |
| firecrawl:scrape | 2026-01-01 | $0 each |




