What we won't do
- No fabricated quotes. If we attribute a statement to a person or an organisation, that statement exists in the cited source. We do not invent expert opinions.
- No fabricated statistics. Every numeric claim links to a primary source we have personally checked. If we link to it, we have read it.
- No AI-generated reviews or testimonials. shrybe does not publish reviews on itself, and we do not run a "what people say" carousel. No
AggregateRatingSchema.org block. No fake user names. - No invented DOIs. When we cite an academic paper, the DOI has been verified at doi.org or directly at the publisher. Models hallucinate DOIs; we don't relay the hallucinations.
- No sponsored content disguised as editorial. If a piece is sponsored or paid, it is labelled clearly. We do not run affiliate links inside neutral comparison content. Currently, none of our pages contain affiliate or sponsored content.
How we research
Our research process is uneventful by design. We work from primary sources whenever a primary source exists. Specifically:
- Academic claims go to the paper itself (preferably the arXiv preprint or the publisher's page). We verify the DOI, the authors, and the year. We quote from the paper, not from a press summary of the paper.
- Statistics go to the original report. If we cite "the AI transcription market was $4.5 billion in 2024," we link to the actual market.us research note that made that estimate — not to a SEO-spam page that reprinted it.
- Standards and specifications go to the standards body. WCAG references go to w3.org/WAI/WCAG22. RFC references go to IETF datatracker. Schema.org references go to schema.org.
- Industry claims are clearly labelled when they come from a vendor's own marketing — Tactiq's "700,000+ users" is a Tactiq self-claim, not an independent audit, and we say so when we cite it.
How we write
Every long-form page on shrybe is written by humans on our team. We do use AI tools as research aids — to draft outlines, to surface candidate sources, to suggest alternative phrasings — but the final text is human-edited and human-vouched for. We treat AI output as a draft, not a fact.
We avoid the language patterns that signal AI-generated content even when factually accurate: empty intensifiers ("revolutionary," "game-changing"), structural cliches ("In today's fast-paced world…"), unbalanced positivity, and the cluster of hedges and repetitions that frontier models default to. The goal is prose that reads like a human wrote it, because one did.
Updates and date stamps
When we update a page in a way that changes meaning — new statistic, revised claim,
corrected error — we update the published date in the page's Schema.org metadata
(dateModified). For substantive rewrites, the change is noted at the bottom
of the page in a small revision line. Cosmetic edits, typo fixes, and link tidying
happen silently.
We are a small team. We do not promise quarterly content audits or weekly refresh cycles. We promise that when we notice an error or a statistic that has aged out, we fix it promptly, and we mark the date so you can see when.
Corrections
If you spot a factual error, a broken link, a misattributed quote, or anything else that looks wrong, please tell us at hello@shrybe.com. We will fix it. For substantive corrections — wrong facts, wrong numbers, wrong attribution — we update the page and add a short "Correction:" line at the bottom describing what changed and when.
We do not silently rewrite content to make ourselves look better in the past. Our older pages get to look like older pages; what we update, we date.
Disclosures
shrybe is bootstrapped. There are no investors, no acquihire pressures, no influencer partnerships, no affiliate revenue. We have no commercial relationship with OpenAI, Anthropic, Google, Tactiq, Otter, Descript, or any of the other companies named on this site. We name them where the comparison helps you, not where someone paid us to mention them.
Our backend uses OpenAI's open-source Whisper model, downloaded and run on our own hardware. We are not an OpenAI partner, reseller, or affiliate — we just use the model. The same goes for the rest of our backend dependencies: open-source, community-maintained, and we contribute issues upstream when we hit edge cases.
Privacy of the content you transcribe
This is covered in detail on the /about and product pages, but worth
repeating in the editorial-standards context: we do not republish, train on, or analyse
the transcripts produced by the free tier. They are delivered to your browser and
deleted on our side within 24 hours. If we ever change that policy — for example, to
support a paid feature that retains transcripts for the user's own re-reading — the
change will be announced in advance and clearly described on the privacy page that will
live at /privacy by then.
If something feels off
Email hello@shrybe.com. We read everything. We do not have a support ticket system because we do not have enough volume to need one. A thoughtful question from a stranger has a real chance of changing what shrybe does.
Standards are easier when the product is honest.
Back to /about, or read about /pricing for the financial side of the same conversation.