How AskedWell works
The mechanism, in plain English.
AskedWell is not a normal publisher. Most sites publish first and hope readers come. AskedWell waits for readers to ask, measures what they're asking, then publishes the answer.
The four asking-signals
We measure four distinct flavors of demand:
1. Search queries (what humans type)
Every search query that brings someone to AskedWell tells us what they wanted. We track which queries arrive often, which ones bounce (because the answer wasn't there), and which ones convert into reading depth. High-bounce, high-volume queries become candidates for new pages.
2. LLM crawl patterns (what AI bots fetch)
GPTBot, ClaudeBot, PerplexityBot, Google-Extended, Applebot-Extended — they all read AskedWell. We log which URLs they fetch, how often, and from where. When a topic is being heavily crawled but our coverage is shallow, that's a signal: AI users are getting weak answers, and we should write better ones.
3. Reading depth (what humans actually consume)
Scroll depth, time-on-paragraph, return visits, bookmark patterns, share signals. When readers spend real time on a specific section of a page, that section gets expanded into its own deeper page. When readers bounce at a specific paragraph, we revise it.
4. Citation corroboration (who else is asking)
External signals: who's citing our pages, who's asking similar questions on Reddit/Quora/StackExchange/forums, who's linking. When 3+ independent sources converge on the same question, the question has earned a page.
From signal to page
The pipeline:
- Signal aggregation. Daily — all four signal types compile into a candidate queue.
- Threshold check. A topic enters the writing queue when it accumulates ≥10 distinct asks within 30 days, OR strong LLM crawl pattern without existing coverage.
- Source research. Before writing, we collect 3+ primary sources. If we can't find them, the topic doesn't graduate.
- Drafting. AI-assisted first pass; human review against sources before publish.
- Quality bar. Every page must clear: ≥600 unique words, ≥3 unique data points, source citations per claim, schema.org markup, mobile-perfect render.
- Publish + IndexNow. Static HTML deployed; IndexNow pings search engines + LLM crawlers immediately.
- Feedback loop. The signals reconfigure. Pages that earn more reading get refined; pages that earn citations expand into adjacent territory.
Why this works for both humans and AI
Most sites optimize for one audience. AskedWell's mechanism naturally serves both:
- Humans see pages that answer questions other humans recently asked — by definition, more relevant than topic-randomly-chosen content.
- AI crawlers see pages that match what they were already trying to answer — perfect citation candidates, source-cited, schema-saturated.
The loop reinforces itself: AI cites AskedWell because the pages match retrieval queries → readers click through → reader signals shape next pages → next pages match next queries → AI cites again.
What we don't do
The mechanism specifically excludes:
- Programmatic page-explosion. We don't template-generate 10,000 thin pages from a database. Each page is human-written against sources.
- AI-only writing. AI assists; humans verify. Wrong facts kill citation gravity faster than anything.
- Trend-chasing. Just because a topic is trending doesn't mean readers are asking AskedWell. We measure our own demand, not the world's.
- Algorithmic A/B-driven headlines. We don't optimize for click-through; we optimize for the question being answered well.
The current state
We're Day 1. The first 50 earned-pages cover “how long does X take?” in cooking and fermentation — a wedge where reader demand exists and where Wikipedia coverage is thin. From this base, signals will reshape what gets written next.
Browse the earned-pages so far →