Use AI on social without sounding like everybody else
AI social media training teaches your team to use generative AI for posts, captions, carousels and short form video without producing the flat, obviously generated content audiences scroll past. One private session, your channels, your brand voice, from £1,200 for up to six people.
Why does AI social content look so obviously AI?
Because almost everybody prompts it the same way. Ask any model for five LinkedIn posts about a topic and it reaches for the same shapes. A rhetorical question opener. Three items in a list, always three. A short sentence for emphasis. Then a question at the end to drive engagement.
Your audience has now seen thousands of those. They recognise the pattern long before they consciously register that a machine wrote it, and the cost is not that they dislike the post. It is that they stop reading yours specifically, because it looks like the wallpaper.
This session is built around fixing that. We spend the day on the two things that actually change the output, which are a brand voice reference built from your own best performing posts, and a working list of the tells that give generated content away.
Does AI generated content get penalised by social media algorithms?
Not for being AI. That is the short answer, and it is not the answer most courses give. What actually happens is more specific, and more awkward, platform by platform.
| Platform | What it actually does | What it means for you |
|---|---|---|
| YouTube | Since 15 July 2025 its policy names inauthentic content. Mass produced, generic or templated material is ineligible for monetisation. | Volume without a point of view is the risk, not the tool that made it |
| TikTok | From November 2025 viewers can dial down how much AI content appears in their feed. Invisible watermarking and C2PA credentials are applied, and TikTok says it has labelled over 1.3 billion videos. | Suppression is chosen by the viewer, so it never shows in your analytics |
| Meta | AI info labels applied from industry signals and self declaration. Reduced distribution attaches to fact checker ratings, not to AI labels. | Labelling itself is not a distribution penalty on Meta |
| In January 2026 Adam Mosseri said labelling everything will become impossible, and the focus shifts to verifying authentic content. | Provenance is becoming the signal, not the label | |
| Displays content credentials on images that carry them. | Your image metadata travels with the file whether you mention it or not |
Sources: YouTube monetisation policies, TikTok Newsroom and Meta. Last reviewed August 2026. Platform policy moves quickly here, so we check it again before every session.
Will people trust your content less if you label it as AI?
Yes, and this is the finding nobody teaches. Controlled experiments run by the Nuremberg Institute for Market Decisions across the United Kingdom, the United States and Germany, with a thousand respondents in each market, tested identical adverts. The only variable was whether the advert was labelled as AI generated.
The labelled version was judged more harshly. Respondents rated it less natural and less useful, reacted more negatively on emotional measures, and were less willing to click. The penalty was worse for traditional product categories and milder for technology.
Ofcom’s Adults’ Media Use and Attitudes report, published 2 April 2026, points the same way. Fifty seven percent of AI aware UK adults say they would trust an AI generated news story less than a human written one, and among people who use AI tools themselves half still say it.
So the question is not whether to be honest. It is how to disclose without it reading as an apology, and which parts of the process are worth disclosing at all. We work through that on your own posts rather than in the abstract.
Do you have to label AI generated posts in the UK?
There is no blanket legal requirement to disclose AI use in UK advertising. The Committee of Advertising Practice has been explicit that the CAP Code is media neutral and contains no AI specific rules, so the same rules apply however the content was made.
Disclosure becomes necessary where the audience would be misled without it, or where AI features prominently in a way consumers would not expect. The line worth remembering, and it is CAP’s own, is that disclosure alone is very unlikely to fix a fundamentally misleading message. Labelling an image that shows a result your product cannot deliver does not make the advert compliant.
We cover where your posts sit against that, what to do about AI generated people and voices, and why using a synthetic version of a real person is a different category of problem entirely. See the ASA and CAP guidance. This is training, not legal advice.
What does the session cover?
Seven blocks, hands on throughout, using your own channels, your own analytics export and your own brand guidelines. Nobody leaves with a folder of generic exercises.
1. Where AI helps and where it hurts
An honest map of the tasks worth handing over, and the ones that cost more time than they save once you have finished editing.
2. Building your brand voice reference
We take your best performing posts and your guidelines and build a reusable reference document that goes into every prompt.
3. The tell list
The markers that give generated content away, from the tricolon opener to the symmetrical carousel and the uncanny stock smooth image.
4. Ideation and planning
Turning a campaign brief into a month of angles, then filtering hard. The value is in what you reject, not what you generate.
5. Image and video generation
What the tools can do now, what content credentials attach to the file, and what happens to that metadata when you upload.
6. Disclosure and platform policy
What to label, what not to, and where the ASA position leaves you. Platform by platform, updated before every session.
7. Your team AI rules
Who approves AI assisted posts, what never goes near a model, and how you log it. A one page working document by the end.
Can AI actually write in your brand voice?
Not from an instruction. Telling a model to write in a friendly, professional and approachable tone gives you the same output everybody else gets, because that instruction carries almost no information about you.
What works is evidence. We build a reference document from your fifteen best performing posts, your guidelines, and a banned list of words and constructions you never want to see. That document goes into every prompt, and the difference in the first draft is immediate and obvious to the room.
Teams usually find the exercise useful on its own terms, because writing down what your voice actually is tends to expose where it has drifted since the guidelines were written.
Is AI assisted content actually performing better?
The best public dataset says yes, with caveats worth stating. Buffer analysed 1.2 million posts across seven platforms from fifteen thousand users who published both AI assisted and non AI content. Median engagement was 5.87 percent for AI assisted against 4.82 percent for the rest.
Three things to hold in mind before quoting that at your board. Buffer sells an AI assistant. AI assisted is not the same as AI generated, and most of those posts were edited by a person. And Buffer’s own reading is that the lift comes from volume and consistency rather than from the writing being better.
Their much larger 2026 engagement study does not repeat the comparison, so there is no refreshed number. We say that on the day rather than pretending the evidence is stronger than it is.
How is it delivered, and what does it cost?
| Format | Length | Group size | Price excluding VAT |
|---|---|---|---|
| Private half day, your offices or online | 3 hours | Up to six | From £1,200 |
| Private full day, your offices or online | 6 hours | Up to six | From £1,800 |
| Larger teams and multi site rollouts | Agreed | Seven or more | Quoted |
Public seats elsewhere run £375 to £525 a head for a day. For a team of six that is £2,250 to £3,150, and everybody spends the day working on somebody else’s examples. We price by the group, which works out around £200 a head, and the work is done on your accounts.
We travel across the United Kingdom. Leeds, Manchester, Birmingham and London are routine. Online sessions run on Teams or Zoom in shorter blocks so nobody sits through six hours of screen.
Who runs the day?
I am Chris Norton and I run most of these sessions myself. I founded Prohibition, the PR, social and content agency behind this site, I have worked in public relations for more than twenty years, and I have been running training since 2009. I co-authored Share This Too, the social media handbook published by the Chartered Institute of Public Relations.
We publish social content for clients every week and we use these tools on that work, which means the session is built on what has held up in public rather than on a demo. Where a specialist is needed alongside me, they come to the room.
Is this the right AI course for your team?
You are here
AI for social media content. Posts, captions, carousels, short form video, brand voice and disclosure.
Microsoft Copilot
Documents, decks, spreadsheets and email inside Microsoft 365. See Microsoft Copilot training.
AI search, GEO and AEO
Getting cited by AI Overviews rather than losing the click. See SEO and AI search training.
What does the day actually look like?
| Time | Block | What comes out of it |
|---|---|---|
| 09:30 | Where AI helps, where it hurts | An agreed list of tasks your team will and will not hand over |
| 10:15 | Building the brand voice reference | A working reference document made from your own posts |
| 11:15 | The tell list, and rewriting against it | Three of your recent posts rewritten so they stop reading as generated |
| 12:15 | Ideation and a month of angles | A filtered content plan, with the rejects and the reasons |
| 13:45 | Image and video, and what travels with the file | Assets made in the room, with the metadata question settled |
| 14:45 | Disclosure, platform policy and the ASA | Your disclosure position, written down |
| 15:30 | Team AI rules and sign off | A one page set of rules the team has agreed out loud |
The half day version keeps blocks one, two, three and seven, which is the combination most teams need first. Full day adds the planning, generation and disclosure work.
Who should attend?
Social and content teams
The people writing the posts, building the carousels and cutting the short form video. This is their day.
Marketing and comms managers
Anyone who signs off what goes out and needs a defensible position on AI to give a board or a client.
Agency account teams
Client facing teams who need speed without putting confidential or embargoed material anywhere near a model.
No previous AI experience is needed. We have run this for teams who use these tools daily and teams who had never opened one, in the same room, and the exercises stretch to both.
AI and social media questions we get asked
How much does AI social media training cost in the UK?
Public courses run around £375 to £525 per person for a day, so a team of six costs £2,250 to £3,150. We charge by the group instead, from £1,200 for a half day and £1,800 for a full day for up to six people, excluding VAT. That works out at roughly £200 to £300 a head, delivered in your office on your own channels.
Do we have to label AI generated social posts in the UK?
There is no blanket legal requirement. The CAP Code is media neutral, so the same advertising rules apply however the content was made. Disclosure is needed where the audience would otherwise be misled, or where AI features prominently and would not be obvious. Disclosure does not fix a misleading claim.
Will the algorithm punish us for using AI?
No platform currently demotes content simply for being AI generated. YouTube demonetises content that is mass produced and generic, TikTok lets viewers filter AI content out of their own feed, and Meta and LinkedIn label it. The commercial risk sits with your audience rather than with the ranking.
How do we stop AI posts sounding obviously AI?
Two things do most of the work. Build a brand voice reference from your own best performing posts and feed it into every prompt, and learn the structural tells, which are the rhetorical question opener, the group of three, the short emphatic sentence and the engagement bait question. Once a team can name them they stop shipping them.
Can AI make images and video we can actually post?
Often yes, and the quality question is no longer the interesting one. The interesting one is what content credentials are embedded in the file, what survives an upload, and what a platform does when it reads them. We cover that with the tools your team already has.
What should never go into an AI tool?
Unannounced news, embargoed material, client confidential information, personal data about identifiable people, and anything under an NDA. We set out a workable rule your team can remember rather than a policy nobody reads.
Do we need paid AI tools or is the free tier enough?
For social content specifically the free tiers get you a long way, and we will say so rather than sell you licences. Where paid tiers earn their keep is longer context, image generation quality and keeping your material out of training data, which matters more for the confidentiality question than the creative one.
Is this the same as your social media training?
No. Our general social media courses cover strategy, channels and measurement. This session assumes you already run social and focuses only on where generative AI fits into producing the content.
Can you train an agency team rather than an in house one?
Yes, and the confidentiality material is more relevant to you, not less. Tell us when you enquire and we will build the session around client work rather than brand work.
How many people can attend?
Six is the sweet spot for hands on work. We will run up to twelve, but past that the practical exercises suffer and you are better served by two sessions.
Tell us what your social team is producing and where it is stalling
Half day £1,200, full day £1,800, up to six people, excluding VAT. Delivered UK wide or online.
Part of our AI training for marketing and communications teams, alongside every social media course we run. Last reviewed August 2026.
