We re-read all sixty Directive Academy guides against our live policies. The publisher self-audit found six recurring error categories: fee-timing wording that cast our own lawful invoicing as a warning sign, worked examples using unregistered DOIs under our real prefix, imprecise standards wording, unsourced superlatives, one mislabelled integrity term, and wording that implied indexing and preservation we do not hold.
Every one is corrected on the live pages. This post shows what the wording said, what was wrong, and what it says now.
What the publisher self-audit checked, and how
A publisher self-audit is a systematic re-reading of your own published guidance against the policies and standards it cites, with each mismatch recorded and corrected rather than quietly overwritten. Ours covered the sixty guides on Directive Academy, read in full, one at a time.
The method was deliberately dull. Each guide was read twice: once for what it claimed, once for what it implied. Every factual sentence had to trace to a policy page live on this site, or to a document published by a named body such as COPE, EQUATOR or Crossref. A sentence that traced to neither was rewritten or removed.
The four questions asked of every guide
- Does this describe a policy we actually publish? If the guide promised a rule, that rule had to exist on a live page. Several did not.
- Would this sentence survive a hostile reader? Anything that could be read as claiming an index, a metric or an archive we do not hold was flagged, whether or not it said so outright.
- Is every identifier in the worked examples real? DOIs, ORCID iDs, registry numbers and accession numbers were resolved, not eyeballed.
- Does any frequency claim have a source? Phrases such as "the most common reason" were treated as errors unless a published figure sat behind them.
Seven representative corrections, before and after
Rows are not tagged to individual guides; several of these phrasings recurred in more than one place. The last two rows share one category. The right-hand column is live now.
| What it said | What was wrong | What it says now |
|---|---|---|
| "Being asked for money after acceptance is a red flag." | Invoicing on acceptance is normal, legitimate practice — and it is our own model. The sentence described this publisher as a warning sign. | "Charges must be published before you submit. The warning sign is a fee that is undisclosed, or that is presented as buying a decision." |
| A worked DOI example in the form 10.52338/[journal].2026.[number] | An invented identifier under our live Crossref prefix. It resolves to nothing, and a string that looks registered but is not is what we tell authors to distrust. | Crossref's own documented test identifier, 10.5555/12345678. |
| "Follow the EQUATOR guideline for your design and state which one you used." | Our author guidance asks you to match the manuscript to the relevant EQUATOR checklist. It does not require you to name it. The guide promised a rule we do not operate. | "Match your manuscript to the relevant EQUATOR checklist before you submit. Naming the checklist in the methods is useful to reviewers, not mandatory here." |
| "The most common reason manuscripts are rejected is weak methods reporting." | A frequency claim with nothing behind it. We hold no rejection-reason breakdown. | "Methods a reader cannot follow step by step give a reviewer nothing to assess." The mechanism is named; the rate is not. |
| "A similarity score above 20% means plagiarism." | Two errors in one sentence. A similarity index measures matched text, not misconduct, and the threshold was invented. | "Similarity screening reports matched text, including correctly quoted and cited matches. A score is a pointer for human reading, not a verdict. No percentage separates acceptable overlap from misconduct." |
| "Your article is preserved in a dark archive." | Reads as enrolment in a named preservation service. We are not enrolled in any named preservation archive. | Preservation is described generically: what we hold, in what formats, and which memberships we have not taken out. |
| "Indexed and discoverable in the major databases." | Vague enough to be read as PubMed or Scopus. We are in neither. | "Records are in OpenAlex and Google Scholar. JATS XML and OAI-PMH are available. We are not in PubMed, PubMed Central, MEDLINE, Scopus, Web of Science or DOAJ." |
The correction that was hardest to publish: we described our own billing as a warning sign
The predatory-publishing guidance carried a line saying that a journal asking for money after acceptance is a red flag. That sentence was wrong twice over.
It was wrong as a general claim. Invoicing an article processing charge once a manuscript is accepted after peer review is ordinary practice across open-access publishing. What distinguishes an exploitative charge is not when the invoice arrives, but whether the amount was published before you submitted and whether the money is tied to the editorial decision.
It was also wrong about us. Submission to Directive Publications is free. The APC becomes payable only if and when a manuscript is formally accepted after peer review, and nothing is owed if it is rejected. A guide written to help authors spot bad actors described that arrangement as the signal of a bad actor.
The corrected wording moves the test from timing to disclosure and independence. The charges are on Publication Charges, the three waiver routes are on Fee Waivers & Discounts, and waiver requests are handled administratively, apart from the editorial evaluation. Our guide to identifying a predatory journal now carries the corrected test.
Why we publish categories rather than a total error count
A number would be easier to read and impossible to defend. One faulty phrase recurred across several guides: fixing it is a single decision but a separate edit in every place it appeared, so the same audit could honestly be reported as one error or as many. Counting a rewrite for tone alongside a fabricated identifier would flatten a serious problem into a tidy statistic.
So this post reports categories, with real examples in each, and no total — the standard we apply to any operational figure. If we cannot source it, we do not print it — which is why you will find no acceptance rate, no turnaround time and no impact factor, CiteScore or h-index anywhere on this site.
How a blog correction differs from a correction to the published record
Blog guidance is editorial content without a DOI. When it is wrong, the page is edited and the corrected text stands; that is what happened here. A published article is different. It carries a DOI registered with Crossref, it may already have been read and cited in the form it was published, and it cannot be silently altered. Changes to an article go through the formal route set out on our Corrections Policy page, so a change to the published record is recorded and citable rather than made silently. A publisher that edits articles in place is telling you something.
What changed in the process so the same errors do not recur
- Identifiers must resolve. Every DOI, ORCID iD and registry number in an example is resolved before publication or replaced with a documented test identifier. Examples under our own prefix are real, registered DOIs.
- Policy claims cite the page. If a guide says we do something, it links the policy page that says so. If the policy does not exist, the guide says so plainly instead of describing one.
- Frequency words need a source. "Most", "usually" and "typically" come out unless a published figure supports them, and the mechanism is described instead.
- Standards are named precisely. CONSORT for randomised trials, PRISMA for systematic reviews, STROBE for observational studies, CARE for case reports — with matching a checklist to a design left as the author's judgement, as our guide to CONSORT, PRISMA and STROBE explains.
- Integrity terms follow their sources. Similarity is not plagiarism, a correction is not a retraction, and text recycling is not automatically misconduct.
What the audit did not fix
Three gaps remain. Editorial board verification through ORCID, ROR and OpenAlex is in progress, not complete, and we will not describe that work as finished before it is. We remain unenrolled in any named preservation archive; the audit removed the implication that we are enrolled, but deleting a wrong sentence does not close the gap behind it. And the audit covered blog guidance only, not published articles, which sit under the corrections route described above.
None of this makes the corrected guides authoritative by default. To test the claims rather than take them, the sources are the policy pages themselves — Author Guidelines, Peer Review, Publication Charges and Preservation Policy — and the guides worth reading next are how to identify a predatory journal, how similarity screening works and where image adjustment ends and manipulation begins. Where any guide still contradicts a policy page, the policy page is correct. Tell us, and we will publish the correction the same way we published these.