Editorial Policy

Sourcing

Every factual claim has a source. Pricing comes from vendor pricing pages or dated third-party reporting, cited by name. Capability claims come from vendor documentation or hands-on use.

We do not invent statistics. If we do not have a number, we say we do not have it. This is why our research section currently publishes a pricing synthesis and a measurement protocol rather than primary survey findings: we have no primary dataset, and publishing an invented one would be worse than publishing a method.

We distinguish absence of evidence from evidence of absence. In our comparison matrix, "not documented" means we could not verify something, not that it is missing. Conflating those two is the easiest way to unfairly damage a competitor, and we treat it as a serious error rather than a rounding convention.

We date everything. Published and last-updated dates on every page, and a verification date on the tools dataset. AEO pricing and features change fast enough that undated claims are worthless.

Independence

Ratings are not for sale. No vendor can buy a rating, a ranking, or a position. If we ever run advertising or affiliate links, that will be disclosed on the page and in disclosures, and it will not touch ratings.

We do not accept vendor edits to assessments. Vendors may correct facts. They may not revise our conclusions.

We publish reasoning, not verdicts. Where we recommend something, the argument is on the page so you can contest it specifically. A recommendation you cannot interrogate is an advertisement.

We name competitor strengths. Every competitor review here identifies buyers for whom that tool is the better choice than our editor's choice. A comparison that finds nothing good in the alternatives is not a comparison.

What we will not publish

Fabricated statistics, survey results, or research findings.
Case studies we cannot verify.
Competitor weaknesses we have not confirmed.
Sponsored content presented as editorial.
Claims that a tool can guarantee AI citations or control what a model says. Nothing can.
Content that reads as though it were written to be cited rather than to be read.

That last one is a genuine risk on a site about AEO, and we hold ourselves to it deliberately. Writing to be quoted rather than to be useful produces exactly the hollow, over-structured content this discipline is often accused of encouraging.

Corrections

We correct visibly. Errors of fact are fixed, the last-updated date changes, and where the correction is material we note what changed and when. We do not silently edit.

Anyone can request a correction, including vendors we have assessed unfavourably. Send it through submissions with the specific claim and the evidence. We verify independently.

We distinguish corrections from disagreements. A wrong price is a correction. Objecting to a rating is a disagreement, and we will engage with it but will not change a conclusion because a vendor dislikes it.

Guest contributions

We accept submitted articles for review. Standards for those:

Original. Not published elsewhere, not AI-generated filler.
Sourced. Claims backed by evidence. Statistics with provenance.
Practitioner-grounded. Written by someone who has done the work.
Not a pitch. Vendor content is welcome as clearly labelled vendor perspective, not disguised as neutral analysis.

We edit for clarity and house style, we fact-check claims, and we may decline without detailed feedback given volume. Accepted pieces carry the author's byline and any relevant affiliation disclosed inline.

Submit through our submissions page.

AI use and automated publishing

Some articles are researched, drafted, checked and published through an automated AI workflow without pre-publication human review. Those articles say so on the page and list the primary sources used. Humans set this editorial policy, the site position and the correction standard; automation does not make a source independent or a conclusion proven.

Every automated run is required to check the existing catalogue for duplication, use current primary or authoritative sources for factual claims, distinguish documented facts from editorial inference, avoid invented statistics and guarantees, validate the content build, and verify the published URL. If a source or publication step cannot be verified, the run stops instead of filling the gap with a guess.

We think stating this plainly is more useful than claiming purity. A site about how AI systems consume content that hid its own automated publishing would be an odd thing.

Automation raises the risk of a fluent error reaching readers. We address that risk with visible source lists, dated verification notes, strict limits on unsupported claims, and public corrections. It does not eliminate the risk. Readers should check consequential decisions against the linked primary sources.

Reader contact

Corrections, disagreements, tips and data: submissions.

We would particularly like to hear from practitioners with real contracted pricing data or completed measurement studies, including studies whose findings contradict ours. We will publish methodologically sound work whether or not its conclusions suit us.

Last reviewed 14 September 2026.