how this site's content actually gets made.
this page explains, plainly, how archive's blog is written, reviewed, fact-checked and corrected — including the parts written with ai assistance.
updated: 3 september 2026two kinds of content
this site publishes two different kinds of posts. ordinary how-to and opinion posts (e.g. "how to start a voice journal") describe archive's own product or general practices and carry lower factual risk. comparison, roundup and review posts that name specific competitor products (e.g. "archive vs. day one") carry real factual and reputational risk if a claim is wrong, and are held to a stricter standard described below.
how ordinary posts are made
some posts are drafted with ai assistance against a fixed brief describing archive's real, current feature set — the drafting tool is instructed never to claim a feature archive doesn't have, and never to invent a statistic or study. every generated draft is written with a machine-readable status (draft, reviewed, published, or needs update) and is not published to the live site until it reaches "published" status.
how comparison posts are made — the strict rule
a post naming a specific competitor product is never allowed to auto-publish, regardless of how "clean" an automated check reports it. archive's own automated drafting pipeline is deliberately built to refuse naming competitors at all, precisely because it has no live way to verify anything about another company's current product. every comparison post that does name a competitor on this site today was written and fact-checked by a human, directly against that competitor's own official website, official app store listing, or official documentation — never a third-party roundup, a search snippet, a forum post or an ai-generated summary used as the sole source.
what a comparison post is required to show
every comparison/roundup post on this site carries a visible methodology section: which products were included and why, an explicit disclosure when archive itself is included ("archive is our own product"), a "last checked" date distinct from when the page copy last changed, and a list of the actual official sources consulted, linked directly. subjective claims ("the clear leader," "the best," "most complete") are only used when the surrounding text states the specific criteria behind the claim — never as an unsupported assertion.
ai's role, stated plainly
ai assists with drafting and with a first review pass (checking a draft against archive's real feature list, flagging tone drift, catching invented statistics). ai does not replace fact-checking a claim about a named competitor product — that step is done by a person against an official source, every time, with no exception for a post an automated pass rated "clean."
corrections
if something on this site is wrong — a stale price, a feature that changed, a genuine error — email us with the specific page and claim. corrected posts get a new "last checked" or "last updated" date reflecting the fix; we don't silently edit a claim without updating the date next to it.
what we won't publish
no claim that archive is live on google play or the app store before it actually is. no claim that archive is "fully private," "fully local" or that audio "never leaves your device" — transcription sends audio to a processing service when you request it, and the site says so. no invented user counts, testimonials, awards, or unsupported claims about journaling and mental health outcomes without a credible, cited primary source.