How to read a comparison page, including ours

Search for "best AI branding tool" and every result on the first page was written by somebody who sells one, or by somebody paid a commission when you buy one. That is not a scandal; it is the structure of the category. The pages exist because the query is commercially valuable, and nobody with no stake in the answer has a reason to spend a week researching it.

We publish fourteen of these pages. So this is not a neutral post, and pretending otherwise would be the first thing on the list below. What we can offer instead is the view from inside: the specific mechanisms by which these pages mislead, including the ones we have to actively work against in our own.

1. Find the row where they lose

This is the fastest test and it takes ten seconds.

Scan the comparison table for a row where the other tool wins. Not a hedge — an actual capability the competitor has and the publisher does not.

If there is no such row, you are reading an advertisement. It does not matter how detailed the table is; a twenty-row table where one side wins twenty times is less informative than a three-row table where the sides split. Real tools have real trade-offs, and a page that found none did not look.

Our own rule here is mechanical rather than editorial: a comparison record on this site will not build unless it names at least two things the other tool does better, and the check runs in CI. We wrote it that way because "remember to be fair" is not a policy, it is an intention, and intentions lose to deadlines. The rule is one of several written up on our editorial standard page, each with the file that enforces it named.

2. Check the date, then check what the date is attached to

Comparison pages decay faster than almost any other content type, because the subject is a company that changes its product without telling the person who wrote about it.

So look for a date. Then ask the harder question: is the date attached to the claims, or to the page? "Updated March 2026" at the top of an article usually means somebody changed a sentence in March. It does not mean the feature table was re-checked against the competitor's live site.

On our pages the date is per competitor, not per site, because eight pages re-verified on different days should not all claim the same date. It also expires: our build fails once a comparison's claims pass 120 days without being re-read. That number is arbitrary but the mechanism is not — the failure mode we were guarding against is a "last verified" line that quietly becomes three years old, which is worse than no date at all, because it converts an honest signal into a false one.

3. Distinguish "they do not have this" from "we did not find it"

This is the mechanism that does the most damage, and it is almost always invisible.

A feature table has a cell. The publisher could not find out whether the competitor does that thing. What goes in the cell?

Put a cross, and you have asserted something false about a real business. Leave it blank, and every reader reads a blank as a cross. Both options are wrong, and most pages pick one silently.

We landed on a third: those cells read "varies / check their site", and the page links out. It is uglier and it is the honest answer. When you are reading somebody else's page, look for whether they have any third option at all. A table rendered entirely in ticks and crosses is claiming a completeness that nobody researching a competitor from the outside actually has.

4. Treat every price as fiction

Prices change constantly, are regionally different, differ between annual and monthly, and are frequently negotiated for anyone above a certain size. A price copied into an article is wrong within a quarter, and the reader has no way to know when it went wrong.

We do not publish pricing figures for anyone, including ourselves, on the alternatives pages. Those pages describe pricing models — "published self-serve tiers", "sales-led, no public pricing" — which is the part that stays true and is also the part that actually affects your decision. Whether a tool is £19 or £29 rarely changes anything. Whether you have to talk to a salesperson always does.

When you see a specific number on a comparison page, check it against the vendor's own pricing page before it affects your decision. It is thirty seconds and it is wrong surprisingly often.

5. Ask what the page wants you to conclude, then read for the opposite

Every comparison page has a preferred conclusion. That is fine. The question is whether the page is strong enough to survive containing the argument against itself.

The section worth looking for is the one that says when you should not buy from the publisher. On our comparison pages that is a "when they are the better choice" section, and on the alternatives pages it is a mandatory "stay where you are" case that the record will not build without. If you cannot find that section on a page, you are reading marketing that has been formatted as research.

6. Notice what is not being compared

The strongest move available to a comparison writer is not lying about a row. It is choosing which rows exist.

A tool that is far better at one thing will build a table where that thing is four separate rows. A tool that loses badly on a dimension simply will not have a row for it. Nothing on the page is false and the conclusion is still manufactured.

The defence is to write your own list of criteria before you read anybody's comparison, from your own situation rather than from the categories the industry hands you. Then use their pages to fill your table in, not theirs. This one change does more than every other item on this list combined, and it is the one nobody does, because reading somebody's finished table is much easier than making your own.

The ways ours could still mislead you

A post like this owes you this section, so here it is, specifically.

We chose which competitors get a page. Fourteen exist; there are more than fourteen tools in this category. We picked ones people actually search for alongside us, which correlates with — though is not the same as — ones where we compare favourably. There is no rule in our build that forces us to write a page about a tool that beats us, and that is a real gap in the standard, not a technicality.

We wrote the "when they are better" sections ourselves. They are honest, and they are also written by someone who knows which concessions are cheap. A competitor's own version of that section would name different things.

Our claims about ourselves are not verified by anyone. The rules on our editorial standard constrain how we describe other tools — sourcing, dates, the unsourced convention, no pricing figures. The claims about Repic have no equivalent external check. Where we can make them checkable we have: the brand kits on our example pages are real engine output with the brief printed beside them, and the free tools run without an account so you can test the claim rather than take it. But "our brand model is deeper" is a claim about us, made by us.

The category vocabulary is ours to define, and that is an advantage. When we write that other tools lack a "described brand", we are using a frame that suits us. It is a real distinction and we believe it matters. It is also the frame in which we win, and you should hold it at arm's length for exactly that reason. Six branding words used to avoid a decision is our own argument about vague vocabulary; apply it to us too.

The honest summary is that a vendor comparison is a useful input and a bad decision procedure. Read four of them, from four vendors, note where they disagree, and go and test the two that survive. The disagreements are where the real information is — and no single page, including this one, can show you those.

The companion piece is what it actually costs to leave a branding tool: comparison pages make switching look free, and the six costs that never appear on them are what actually decide whether a move is worth making.

Two follow-ups apply the method to our own pages rather than to anybody else's. Seven tools that beat us is this site's comparison data with the flattering half removed, and why we publish nobody's prices explains the one column we leave out of every comparison and alternatives page, including what that omission costs you on the day you are trying to work out whether you can afford something.