Should I cite you or the study you are quoting?

The study, in almost every case. Where we report someone else's finding, they did the work and the primary source is what a reader needs to check. Cite this site only for something that originates here — our own reading of a body of literature, our comparison data, or an argument we are making.

The organisation. Pages here carry a team byline rather than an individual one because no named person has taken editorial responsibility for them yet, and inventing an author to look more citable is the exact practice we criticise elsewhere. Treat HumanFlow as a corporate author.

What date should I use?

The Last reviewed date printed on the page you are citing, which is also the date in the page's structured data and in our sitemap. Those three agree by build-time check rather than by intention.

Do I need permission to quote a page?

No. Normal quotation with attribution needs no permission from us. If you want to reproduce a whole page or a table wholesale, ask — not because we will refuse, but because we would rather point you at a version that will not go stale.

Will the page still be there?

We do not delete pages. Where something is superseded we correct it in visible text and record it in the corrections log rather than editing quietly, so a citation to a page here should keep resolving to something that acknowledges what changed.

Can I cite a figure you quote from a vendor?

Cite the vendor's page, which we link on every figure along with the date we read it. Our value there is the capture date, not the number — and if you need the number for anything serious you should read it from the source yourself, because vendor pricing and limits move.

How to cite this site

Cite HumanFlow as a corporate author, dated to the page's visible review date. Where we are reporting somebody else's research, cite them instead — we link the primary source on every figure precisely so you can go past us.

Last reviewed 15 August 2026

Cite the source, not the summary

This is the part worth reading before the formats. A large share of what is on this site is a report of work done elsewhere — the studies in the literature review, the vendor figures on the comparison pages, the university decisions in our answers. In all of those cases the citation a reader needs is the original.

There is a practical reason as well as a scholarly one. A summary can be wrong, and ours has been: we have corrected a miscitation on this site where a paper was quoted at half length in a way that changed what it found. A reference chain that runs through a secondary source inherits that risk, and the primary source is one click away on every page.

Cite us for the things that originate here. An argument we are making, our reading of what a body of research adds up to, the dated capture of a vendor's published figures, or a piece of our own analysis.

Formats

Using the glossary entry on perplexity as the example. Substitute the title, URL and review date of the page you are actually citing.

APA (7th)

HumanFlow. (2026, August 15). Perplexity: what it measures in AI detection. https://www.humanflow.pro/glossary/perplexity

MLA (9th)

"Perplexity: What It Measures in AI Detection." HumanFlow, 15 Aug. 2026, www.humanflow.pro/glossary/perplexity.

Chicago (notes)

"Perplexity: What It Measures in AI Detection," HumanFlow, last reviewed August 15, 2026, https://www.humanflow.pro/glossary/perplexity.

Harvard

HumanFlow (2026) Perplexity: what it measures in AI detection. Available at: https://www.humanflow.pro/glossary/perplexity (Accessed: 15 August 2026).

The four fields, and where to find them

Author. HumanFlow, as a corporate author. Pages here carry a team byline because no individual has taken named editorial responsibility for them, and we would rather say that than invent a person with credentials to look more authoritative. Our position on this is set out on the methodology page.

Date. The Last reviewed line printed near the top of the page. That same date is what appears in the page's structured data and in the sitemap, and a build-time check fails if the three disagree — so the date you can see is the date the machines see.

Title. The page's heading, not the browser tab. The two differ deliberately on most pages here.

URL. The canonical one, always beginning https://www.humanflow.pro with no trailing slash and no query parameters. It is in the page source as the canonical link if you need to confirm it.

What we will not do to be more citable

Invent an author, publish a benchmark we have not run, or put a number on something we have not measured. Those are the three shortcuts that make a page look more authoritative to a reader skimming for a citation, and each of them makes it less trustworthy to a reader who checks.

Where a figure could not be established, our pages say so in words rather than filling the gap. Where we have got something wrong, it is recorded in the corrections log with the date and what the page previously said — including one correction that makes our own product look worse. If you are deciding whether this site is worth citing, that log is a more useful thing to read than this page.

If you find an error

Tell us and we will fix it in visible text on the page and record it in the log. That is the arrangement we would want from a source we were citing, and it is the only one that makes a correction useful to anyone who already quoted the old version.

Common questions

Should I cite you or the study you are quoting?: The study, in almost every case. Where we report someone else's finding, they did the work and the primary source is what a reader needs to check. Cite this site only for something that originates here — our own reading of a body of literature, our comparison data, or an argument we are making.
Who is the author?: The organisation. Pages here carry a team byline rather than an individual one because no named person has taken editorial responsibility for them yet, and inventing an author to look more citable is the exact practice we criticise elsewhere. Treat HumanFlow as a corporate author.
What date should I use?: The Last reviewed date printed on the page you are citing, which is also the date in the page's structured data and in our sitemap. Those three agree by build-time check rather than by intention.
Do I need permission to quote a page?: No. Normal quotation with attribution needs no permission from us. If you want to reproduce a whole page or a table wholesale, ask — not because we will refuse, but because we would rather point you at a version that will not go stale.
Will the page still be there?: We do not delete pages. Where something is superseded we correct it in visible text and record it in the corrections log rather than editing quietly, so a citation to a page here should keep resolving to something that acknowledges what changed.
Can I cite a figure you quote from a vendor?: Cite the vendor's page, which we link on every figure along with the date we read it. Our value there is the capture date, not the number — and if you need the number for anything serious you should read it from the source yourself, because vendor pricing and limits move.