Naming a competitor is a commercial decision, not a writing one

A comparison page is the most commercially valuable thing a B2B blog can publish, and the one most likely to get you into trouble. It is also the single format where a generated draft will hand you a confident, well-formed sentence about a company you do not control. Whether to name a competitor is not an editorial decision. It is commercial, it carries a legal edge, and it has to be settled before the engine writes anything.

The comparison page is worth the risk it carries

The safest rule is to never name anyone. Plenty of companies run that rule, and it costs them more than they notice. Buyers late in a decision do not search for your category — they search for a competitor's name with "alternative" or "vs" attached, and somebody is going to answer that query. If it is not you, it is an affiliate site with a scoring table and a commission.

These are also the pages a sales team actually uses. We have argued that the fastest measurable return from a young archive is the post a salesperson sends after a call, and the comparison page is that post more often than any other. A blanket ban on naming competitors is not a cautious policy. It is a decision to be absent from the highest-intent query in your market.

The engine's picture of a competitor is a memory, not a source

Ask a language model about a competitor and you will get a fluent paragraph. Some of it will be right. The parts that are wrong will read exactly like the parts that are right, which is why our sourcing rule is structural rather than a matter of confidence: a claim about another company is a claim about the world, not a claim about you, and it needs a named primary source.

There is a second problem that sourcing alone does not solve. A claim checked against a competitor's own pricing page this morning has a shelf life measured in weeks, because they can change that page without telling anyone, and your post will go on asserting the old number in the present tense. A stale claim about your own product is embarrassing. A stale claim about theirs is the kind somebody's lawyer writes a letter about.

"Verifiable" is a legal word before it is an editorial one

Comparative advertising is legal across the EU, and it is legal conditionally. Directive 2006/114/EC sets out the conditions in Article 4: a comparison must not be misleading, it must compare goods or services meeting the same needs, and it must objectively compare features that are material, relevant, verifiable and representative. It must not discredit or denigrate the competitor.

Read that as an editorial brief and it is unusually precise about the failure mode. Verifiable is the load-bearing word, and it is exactly the property a confident generated sentence lacks. "Their onboarding is slower" is not verifiable. "Their published onboarding timeline is six weeks against our two" is, for as long as the page you took it from still says so.

Sort the claim by who it is about

The rule that survives contact with a real publishing schedule does not ask whether competitors may be mentioned at all. It asks who a given sentence is about, and routes each kind differently.

Claims about you come from the brief, where you have already decided what you are willing to say. Claims about the category the engine can write unaided, because "tools in this space usually bill per seat" commits nobody and is checkable against the market. Claims attached to a named company need a dated quotation from that company's own public page, carried into the draft with its source, or they do not get written. The refusal is the feature. An engine that comes back with "no defensible source for this comparison" has done better work than one that fills the gap smoothly.

It also changes where the page lives. A comparison is a current fact, and a current fact belongs on a maintained page rather than a dated post, because a post asserting a fact ages into a lie while nobody is looking. Give it an owner and a review date, or leave it unpublished.

What the engine cannot decide for you

Two judgments stay with a human, and neither of them is about writing.

The first is defensibility. An engine can tell you a sentence is sourced. It cannot tell you whether you want to stand behind it in front of that company's lawyer, or in front of their sales team, who will read the page within a week of it going live. Sourced and defensible are different tests, and only one of them automates.

The second is whether to name them at all, which is a positioning question. A challenger who names the market leader borrows their authority and usually gains from the trade. A leader who names a challenger hands over attention they would not otherwise have received. Nothing in the draft tells you which of the two you are. The risk was never machine authorship — Google's objection is content nobody needed, not how it was produced. The risk is a fluent sentence you cannot defend.

Defensibility is the test, not permission

Accept this and the question at brief time changes shape. It stops being "are we allowed to mention competitors" and becomes "which claims about a named company are we prepared to defend, and where does each one come from". A founder can answer that in twenty minutes, once, and it converts into a rule the engine applies every day without asking again.

Write the rule down before the first comparison page. Not after the first letter.