About AnswerOpen and the people behind it
Key takeaways
- Built by Mike Hsu, an independent engineer, after auditing two of his own production sites.
- The check catalog comes from published studies and primary operator documentation, with untraceable numbers deleted rather than repeated.
- No testimonials, no customer count, no logo wall, because there are no customers yet.
- We never cloak: fixes go to your origin, visible to humans and engines alike.
- The product name is currently being decided by a pre-registered paid-search test, which is why this site is generated from one brand configuration file.
AnswerOpen is a small, independent engineering product built by Mike Hsu. It exists because an audit of our own production sites turned up crawler and extraction problems that no tool we could buy would have shown us — so we built the auditor, ran it on ourselves first, and fixed what it found.
Who builds this?
Mike Hsu, founder and principal engineer. The product grew out of operating two live sites — a 679-page tax-resolution site and a 391-page short-stay hospitality site — where AI answer traffic was becoming visible enough to matter and no existing tool explained the mechanism. The engine, the check catalog and the fix pipeline are all built in-house. Contact goes to a person, at smhsu11@gmail.com.
How was the check catalog built?
From a structured research pass over 2025–2026 measured studies and primary operator documentation, deliberately separating what has been measured from what is being marketed. Sources include the peer-reviewed GEO work presented at KDD 2024, large-sample citation analyses, and the published crawler documentation from OpenAI, Anthropic, Google, Microsoft and Perplexity. Where three independent passes could not trace a widely repeated number to a source, the number was deleted and only the direction kept.
| Decision | Rationale | What it costs |
|---|---|---|
| Schema weighted at 8 of 100 | No measured citation lift from adding JSON-LD in a matched 1,885-page analysis | A less impressive-looking checklist than competitors who sell schema as a growth lever |
| llms.txt scored at zero | 97% of llms.txt files get no bot requests across 137,210 domains | We decline a feature every rival advertises |
| Off-site factors reported, never scored | Mentions correlate about three times as strongly as backlinks; roughly 82% of citations are earned media | Our score can never claim to be the whole picture |
| No headless rendering | No major AI crawler executes JavaScript, so rendering would measure a page no engine sees | We cannot audit what a JavaScript app shows a human — which is the point |
What will this company not do?
- No invented social proof. There are no testimonials on this site, no customer count and no logo wall, because we have no customers yet. The two sites in our numbers are our own and are labelled as such on every page they appear.
- No promised outcomes. Nobody can guarantee a citation from an answer engine, and any product that implies otherwise will be falsified by your own dashboard within a quarter.
- No cloaking. We never serve engines a different copy of a page from the one humans get. Fixes are applied to your origin, visible to everyone.
- No untraceable statistics. Every number on this site traces to a source we can name, and the ones about our own sites come from consecutive runs of the same engine you would be buying.
Architectural honesty is the moat. If our fetcher cannot see it, neither can ChatGPT — so we skip headless rendering entirely, and a ten-thousand-page crawl stays cheap enough to re-run weekly.
— AnswerOpen product notes, August 2026
Why might the name change?
Because it is being tested rather than assumed. The name is currently the subject of a live paid-search test against a second candidate, with the decision rule written down before any data arrived: identical keywords, bids, geography and ad copy, with only the brand token differing, read out on click-through rate with a two-proportion z-test at 2,000 impressions per arm. The site you are reading is generated from a single brand configuration file for exactly that reason — if the other name wins, one config value changes and the whole site rebuilds under it. The product does not change either way.
More: how an audit runs · the evidence behind the checks · get in touch.
Primary sources
Everything asserted above traces to one of these. Operator documentation changes often; check the current version before relying on any of it.
- Generative Engine Optimization (KDD 2024) — the only peer-reviewed causal evidence on content modifications and generative visibility
- OpenAI crawler documentation — the authoritative list of GPTBot, OAI-SearchBot and ChatGPT-User behavior
- Anthropic crawler documentation — ClaudeBot, Claude-SearchBot and Claude-User
Request an audit of your site
There is no fake "instant scan" button on this site. A real audit crawls your sitemap, fetches every page, and runs nine live user-agent probes against your edge — it takes minutes of machine time, not milliseconds, and it is run by a human who reads the result before you see it. Send the URL and you get the headline score plus your single largest blocker back, at no charge.
Frequently asked questions
Who is behind AnswerOpen?
Mike Hsu, founder and principal engineer. The product came out of operating two live sites, a 679-page tax-resolution site and a 391-page hospitality site, where AI answer traffic mattered and no existing tool explained the mechanism.
How many customers do you have?
None yet, and we are not going to pretend otherwise. The two sites in our published numbers are our own and are labelled that way everywhere they appear.
Where do your statistics come from?
Published large-sample studies and primary operator documentation, including the peer-reviewed GEO work from KDD 2024 and the crawler documentation published by OpenAI, Anthropic, Google, Microsoft and Perplexity. Numbers we could not trace to a source were removed from our catalog.
Why might the product name change?
The name is being decided by a paid-search test with rules written down before any data arrived: identical keywords, bids, geography and copy, with only the brand token differing, read out on click-through rate. This entire website is generated from a single brand configuration file, so the winner is one config change and a rebuild.
Do you resell or publish audit data?
No. A report about your website is yours. We do not publish it, resell it, or name your site as an example without written permission, and we delete our copy on request.