We use AI tools in making this site. This page says how, because a reader deciding whether to trust writing about attention deserves to know how the writing was made.
The alternative — using the tools and staying quiet about it, which is currently the norm — seemed like a poor way to open a relationship built on saying what we actually know.
Our position
AI assists. It does not author, and it does not carry responsibility.
That distinction is doing real work. A tool that helps find a source is not the same as a tool that supplies one. A tool that suggests a structure is not the same as a tool that decides what is true.
What AI is used for
- Research assistance. Finding candidate sources, locating papers, surfacing what has been written about a subject.
- Structure. Drafting outlines, proposing an order of argument, identifying what a piece is missing.
- Checking our reasoning. Arguing against a draft, finding the weak step, asking the obvious objection we had stopped seeing.
- Editing. Tightening, cutting, catching repetition and the sentences that sound better than they mean.
- Production. Formatting, metadata, the mechanical parts of publishing.
All of that is ordinary, and none of it changes who is responsible for the result.
What AI is never used for
Generating citations. Every source referenced on this site has been opened and read by a person. This rule exists because these systems produce plausible references that do not exist — correctly formatted, correctly attributed to real researchers, describing studies nobody ran. A fabricated citation is worse than no citation, because it is credible.
Generating statistics, findings, or quotations. If a number appears here, it came from a source we located and read. If someone is quoted, they said it.
Producing articles at volume. We publish roughly one piece every two to three weeks. That is not a rate that scaled generation would improve, and generating many pages primarily to occupy more search results is a practice search engines classify as abuse. We do not do it for that reason and also because it produces exactly the content this site exists as an alternative to.
Claiming first-hand experience. If an article says we tried something, a person tried it. Where our authority for a claim is observation rather than research, the text says so.
Inventing people. No fabricated authors, no invented credentials, no reviewer names on pieces nobody reviewed. See Who we are, which is conspicuously short on biography for the same reason.
Manipulating search or answer systems. No text addressed to AI systems rather than readers. No structured data describing content that is not on the page. No pages built to be retrieved rather than read. The published guidance from search platforms is clear that these do not work, and we would not want them to.
Fact-checking and verification
Every factual claim on this site goes through the same process regardless of how the draft was produced:
- The source is located — the actual paper, the actual documentation.
- The source is read by a person, at least far enough to confirm it says what the draft claims it says.
- The conditions are recorded — what was measured, on whom, when, and what the study does not establish.
- The claim is labelled as observed, inferred, or a position, using the labels described in the editorial policy.
Step three is where most of the errors in this genre would be caught if anyone performed it. A finding reported without its conditions is not quite false and is not usable either.
Who is responsible
Eonara is. Not a tool, not a workflow, not a process.
If something on this site is wrong, it is our error, and the corrections page sets out how it gets fixed and how the fix is shown.
Why we are telling you this
Two reasons.
The first is that guidance from search platforms treats disclosure as relevant where automation substantially contributes to content, and asks the useful questions: who made this, how, and why. We would rather answer those in public than hope nobody asks.
The second matters more. This site argues that attention is worth protecting and that most content is not worth giving it to. Making that argument while being evasive about how the argument was produced would be an odd position to hold.
What would change this page
If we start using these tools for something not on the list above, this page changes before the practice does, not after it.