Publish on the public wiki

Only new, verifiable knowledge, written to our standard. That is the whole point.

sinapsi.wiki is a verified memory: public wikis that AI assistants crawl every day and query live over MCP, currently more than a hundred live queries a day on top of the crawlers. Every page ties its claims to primary sources, which is exactly why assistants trust and cite what they find here. Sinapsi is not an ad network and not a content farm: it is a memory, and it only grows with content that deserves to be in one.

If you have knowledge worth citing, you can publish it here under your own attribution: when an assistant uses your page, the credit is yours. A product announcement written by the maker is a primary source, and it is welcome, as long as it clears the same bar as everything else.

This is GEO, generative engine optimization, done honestly: instead of gaming answers, you place verified knowledge where AI already reads, and you get the numbers to prove it. Every public wiki has a live measurement panel, see a real example: AI crawler visits, live AI queries, and verified uses, straight from our server logs. A floor, not a promise.

What is GEO, and which methods actually work?

GEO (generative engine optimization) is getting your knowledge found, used and cited by AI assistants: ChatGPT, Claude, Perplexity, Gemini and the rest. Where SEO optimizes for a results page, GEO optimizes for an answer: the assistant either cites you or it does not.

Most of what is sold as GEO is guesswork. The methods that hold up in practice are the boring ones:

• Be a primary source. Assistants prefer the origin of a fact over the tenth rewrite of it. Original data, first-hand announcements, real numbers.
• Tie every claim to a source. Verifiable content is the kind a cautious model is willing to repeat. Unverifiable marketing is the kind it skips.
• Structure for retrieval. Short anchored sections, one claim per anchor, explicit links, dated provenance. Assistants quote chunks, not walls of text.
• Attribute clearly. If your name is not on the claim, the citation credits nobody.
• Be where AI reads: crawlable HTML, machine-readable twins, live query endpoints (MCP). A PDF behind a login does not get cited.
• Measure, don't guess. Crawler logs and live queries tell you what AI actually reads. Everything else is opinion.

This page is the practical version of all of the above: sinapsi.wiki already does the structure, the verification, the machine channels and the measurement. You bring the knowledge.

What gets published

A page goes live only if it passes all five checks:

• New — a search on the wiki does not already return the substance. The test is our own engine, not a reviewer's opinion.
• Proven — every load-bearing claim declares how it is known to be true, and passes the check for that way. See below.
• Citable — there is at least one assertion specific enough to be extracted and attributed: a number with its date, a named mechanism, a procedure, a measured result.
• Well-built — house format: anchored sections, links to related pages, dated provenance.
• Attributed — the author or brand is named clearly, so every citation credits you.

You don't need a link for everything

Demanding a URL for every claim rejects precisely the content worth the most: ten years of hard-won practice is new because it isn't already public. So every claim declares its regime of proof:

• Citational — a public source exists: give the URL and the literal quote that supports the claim. We fetch it and check it.
• Attested — you are the primary source: your data, your telemetry, your practice. We cannot confirm it, and that is not a defect. Give who attests it and by what method, and the claim must be falsifiable in principle by someone with access: "4,200 requests per second measured on 12 March using method X" is an attestation; "our system is fast" is marketing and does not pass.
• Validated — the claim has already held up across verified uses inside Sinapsi. We read that from the ledger ourselves.

A page may mix all three. The one fraud possible here is declaring a source citational when it does not say that thing, and it is the hardest rejection we have. We do not promise every claim is true: we promise you know what kind of evidence it is, and whose name is on it.

What we turn down

Content the wiki already says, claims with no declared regime of proof, cited sources that do not say what is claimed, pages of pure generality with nothing to quote, and anything unlawful or defamatory. Payment never buys a slot: what does not pass the bar does not go up, and that refusal is precisely why being published here is worth something.

How it works

1. Register for free and connect your AI assistant to the Sinapsi MCP with your account token (see Developers). Your assistant reads the rules with the how_to_build_wiki tool.
2. Your assistant runs the audit, at your expense, not ours: for every claim it fetches the source and extracts the quote that supports it, then submits through contribute with the completed self-assessment. Incomplete audits are rejected on the spot.
3. We verify, free of charge. Our own engine re-checks the sources and the novelty, and a human has the last word. We may ask for changes. Free reviews are limited per account to keep the queue honest.
4. If it passes, you pay €9 per verified page beyond your allowance (subscriptions include a monthly number; pages declaring more than 20 sources get a quote) and the content goes live with your attribution, readable by people and queryable by AI assistants over MCP.

The payment link is provided after your content passes review.

Send us your content — free review Or connect your assistant over MCP and let it submit directly. Questions: support@42rows.com

Frequently asked

Will AI assistants actually find my content? The wikis here are crawled daily by the assistants' own bots and queried live over MCP. We can show you the real traffic. What we sell is the review and the publication, not a guaranteed number of citations: no honest vendor can promise those.

Can a company publish about its own products? Yes. The maker is a primary source about its own product. The bar stays the same: new, proven, citable, well-built, attributed. Your own data counts as attested proof, with no external link required.

What if my content does not pass? You get precise corrections: what fails, where, and what it would take. What already holds stays confirmed and does not need rewriting. You only pay after it passes.

Who reviews it? An automated verification engine re-checks every source, and a human makes the final call. The same standard our own wikis are held to.

Honesty notes

Sinapsi is run by Sinapsi The fee buys the evaluation and publication work, never a guaranteed slot or a guaranteed outcome. By submitting you declare that the content is public, that you hold the rights to it, and that you authorize its publication on sinapsi.wiki under CC BY 4.0 with your attribution, like the rest of the public wiki. Questions: support@42rows.com.