Treadstone Associates
Article · Marketing & business development

Can AI write LinkedIn posts for partners?

A partner post promoting the firm is a representation to the public before it is content. Three separate rulebooks apply to it, and a model has read none of your files.

Treadstone Associates · Updated 2026

Key takeaways

The short answer

Yes, for the draft. A model is good at structure, at turning a partner’s spoken thoughts into a readable 200 words, and at adapting one idea into several formats. It is bad at the two things that make a partner post worth reading: a specific opinion held by a named person, and a fact that person can stand behind. Use it for the first and never for the second.

The practical rule that survives review: the partner supplies the position and the evidence, the model supplies the sentences, and the partner reads the result out loud before it goes anywhere. If it does not sound like something they would say to a client, it is not ready.

It is advertising, whatever it looks like

A post that promotes the supply of a service or promotes a business interest is a representation to the public. Section 74.01 reaches a representation that is false or misleading in a material respect, and separately a statement, warranty or guarantee of performance or efficacy that is not based on an adequate and proper test — with the proof of that test on the person who made the representation. Subsection 74.03(5) directs that the general impression be considered alongside the literal meaning, which is exactly where generated copy fails: each sentence is defensible and the paragraph as a whole promises something nobody intended.

The Competition Bureau’s summary of false or misleading representations and deceptive marketing practices sets out the two regimes, and section 74.1 sets the administrative monetary penalties a court may order — for an individual, the greater of $750,000 (and $1,000,000 for each subsequent order) and three times the value of the benefit derived, where that can reasonably be determined. Treadstone Law covers the wider obligation in its overview of misleading advertising rules and of comparative advertising.

Testimonials and the reshared client quote

Section 74.02 is the provision partners trip over. Publishing a testimonial about a service is reviewable conduct unless it was previously made or published by the person who gave it, or was approved and permission to publish it was given in writing before publication — and it must accord with what that person actually said. A model asked to “tighten” a client’s message before you post it has just moved you outside that. Clause 7 of the Canadian Code of Advertising Standards says the same thing from the industry side: a testimonial must reflect the genuine, reasonably current opinion of the person making it. Treadstone Law’s guide to using testimonials and endorsements is the plain-language version.

Platform policy is a separate constraint

LinkedIn’s policies are explicit that members should not spam the platform with untargeted, irrelevant, obviously unwanted or gratuitously repetitive promotional content, should not use the invitation feature to send promotional messages to people they do not know, and should not artificially increase engagement by arranging in advance for others to like or re-share their content. Three AI-enabled tactics are therefore out before you reach any question of law: bulk connection messaging, generated comments at volume, and coordinated engagement.

The regulated-partner layer

If the partner holds a licence, their post is marketing within the meaning of their own rules. Rule 4.2-0 says marketing includes advertisements and other similar communications in various media, and rule 4.2-1 requires marketing to be demonstrably true, accurate and verifiable, neither misleading nor likely to mislead, and consistent with a high standard of professionalism. The commentary lists examples that may contravene it — stating an amount recovered or a degree of success without a statement that past results are not necessarily indicative of future results, suggesting qualitative superiority to other licensees, raising expectations, and testimonials that contain emotional appeals. A model writing engagement-optimised copy produces all four by default. Check your own body’s wording rather than assuming it matches Ontario’s.

Client confidentiality is the other one. “We just closed a tricky matter for a manufacturer in the region” can identify a client in a small market. For an Ontario lawyer, rule 3.3-1 requires information concerning the business and affairs of a client to be held in strict confidence, and the commentary notes the ethical duty is wider than the evidentiary privilege and applies whatever the source of the information.

A drafting workflow that holds up

  • • Partner speaks first. Two or three minutes of dictation containing the actual opinion. The model never originates the position.
  • • Claims register. Any sentence with a number, a comparison or a superlative is listed with its source before the draft is written, and comes out if the source is somebody’s recollection.
  • • No client detail without written permission. Not the sector, not the size, not the outcome, if the combination narrows the field to one.
  • • Read-aloud test. If the partner would not say it to a client, it does not go up.
  • • One reviewer. Someone other than the author checks the post against the marketing rule before scheduling.

Worked example (illustrative)

A twelve-person consulting firm in Calgary wants each of its three partners posting weekly. The first attempt used a generated calendar and lasted five weeks before the partners stopped recognising their own voice.

The second attempt inverts it. Each partner records a short voice note after any client conversation that changed their mind about something. A model turns the note into a draft, keeps the partner’s own phrasing where it exists, and flags every factual assertion for sourcing. Two claims are cut in the first month because nothing supports them; one comparative line about a named competitor is rewritten. Nothing is scheduled without the partner reading it aloud.

Measure conversations started and enquiries that cite a post, not impressions. Impressions are the metric that leads a firm towards the volume tactics the platform policy prohibits.

Where this sits in the firm

This is business development for a firm that sells expertise. Broader campaign mechanics sit under marketing automation. If the underlying question is really about intake, scheduling and reminders at the front desk, that is the professional practice owners page instead.

Questions we get asked

Do we have to disclose that AI helped write it?
No Canadian rule we could find requires it for an ordinary marketing post. What is required is that the content be true and that the named author stand behind it, so the question that matters is whether the partner actually holds the view.

Can we use AI to write comments on other people’s posts?
Generated comments at volume are the behaviour the platform policies describe as spam, and coordinated engagement is called out by name. A partner commenting because they have something to add is a different activity.

What about a post naming a client win?
Get written permission first, and remember that a client quote you edited is no longer the testimonial they gave, which is the point of section 74.02.

Write like the partner, not like the model.

A 30-minute call is enough to tell you whether AI pays for itself here.