Every lawyer knows content marketing works and almost none do it consistently. The pattern is identical everywhere: four posts in two weeks, then nothing for five months.
The cause is structural, not motivational. Treating each post as its own creative act means every post requires a topic, time, and the will to write, and those three rarely coincide on a Tuesday. The fix is to separate creating from publishing, and to do the creating in a single batch.
The method
Once a month, record yourself for 30 minutes answering questions clients actually ask. Just talk. No script, no slides, no production. Your phone is adequate.
Get five to eight questions from your intake team beforehand, taken from real calls that week. Those questions are the content strategy, because they are what people are actually searching, phrased the way people actually phrase it.
Transcribe the recording using any of the tools in AI transcription for lawyers.
Then derive everything from the transcript. From one 30-minute session you can reasonably produce a long article covering the whole session, eight to ten short social posts, one email newsletter, several short video clips if the recording was on camera, and a handful of FAQ entries for your website.
That is a month of publishing from 30 minutes of talking, and the marginal effort per piece approaches zero.
Why this beats writing from scratch
The output contains your actual expertise, because it started as you explaining something you know. The examples are real. The caveats are the ones you actually give clients. That is exactly the material a model cannot invent, and it is what separates content that ranks and converts from content that fills a page.
It also sounds like you. Derived from a transcript, the phrasing carries your rhythm and the words you actually use. The reason so much firm content reads as machine-written is that it was generated from nothing, so it came out as the average of everything.
And it removes the weekly decision. Publishing becomes a scheduling task rather than a creative one, which is why it survives a busy month.
The derivation prompts
Be specific about each output rather than asking for “social posts.”
For the article: turn this transcript into an article that answers the question directly in the first paragraph, uses the speaker’s own examples, and keeps the conversational tone. Do not add information that is not in the transcript.
That last instruction matters. Without it, models pad with generic legal background, and the padding is the part that reads as filler.
For short posts: pull the eight most useful standalone points, each as a short post that makes one point and states it plainly. No hashtag walls, no engagement bait.
For the newsletter: one idea, under 200 words, ending with a single next step.
For FAQ entries: extract every question answered and give a direct two to three sentence answer.
Making it survive review
Read everything before it publishes. Not for tone, for accuracy. The risk is a derived post that states a legal proposition slightly more broadly than you did on the recording, which is the same subtle-overstatement failure that shows up everywhere else AI touches legal work.
Add a jurisdiction note where it matters. A statement true in California and wrong in Texas is a problem when your content reaches both.
Keep the disclaimer visible. General information, not legal advice, no attorney-client relationship. That is standard and it belongs on public content.
The quality bar that keeps this working
The honest test for any piece: could someone who has never handled a case have written this?
If yes, it is filler and it will not perform, because it competes with a thousand identical pages. If it contains a number from your practice, a mistake you watched a client make, or a distinction that only matters if you have actually litigated the issue, it will perform. That is the difference search guidance is aimed at, and the mechanics are covered in AI SEO for law firms.
If you record on camera, Descript handles the clip extraction from the same source file.
Do this today
Ask whoever answers your phones for the five questions they heard most this week.
Then record yourself answering them. Thirty minutes, one take, no editing. That single file is the input for a month of content, and the whole system starts with a question your intake team can answer in two minutes.