AI for marketing and social media, without sounding like a robot

In short

AI multiplies the quantity of content, but audiences reward specificity: real experience, opinions, concrete examples.
Use it to plan, vary and adapt, not to invent what to say.
Build a voice profile and a list of your own topics first: everything else follows.
By law you must disclose deepfakes, i.e. images, audio or video that look authentic and depict real people, places or events (AI Act, Art. 50). For everything else disclosure is a transparency choice, which I recommend.

The paradox of infinite content

Anyone can now publish three posts a day with AI. The result is a sea of similar content, and algorithms (and people) have learned to ignore it. I cover it in the opinion on AI slop: when everyone can produce everything, those who have something to say and say it recognisably win.

So the question isn’t “how do I generate more posts” but “how do I turn what I know and live into useful content, faster”.

Step 1: the raw material is yours

Before opening ChatGPT, list 20 things only you can tell:

questions clients always ask you;
mistakes you see repeated;
concrete (anonymised) cases and how you solved them;
opinions where you disagree with your industry;
numbers and data from your work.

This list is worth more than any prompt.

Step 2: the voice profile

Paste 5-10 texts of yours that work (posts, emails, articles) and ask for a style guide. Save it in a “Communication” project’s instructions: every piece will start from there.

Step 3: the content calendar

With the voice profile saved, ask for a four-week plan: themes tied to your real work, one goal per piece, the channel and the format. You correct it and keep it in the project, so the assistant follows it every time.

Step 4: one idea, five formats

The most effective trick is repurposing: one strong idea becomes an article, a carousel, a short-video script, a post and a newsletter paragraph. AI is great at adapting, much less at creating from nothing.

Images and video

Graphics for carousels and posts: Canva or Gamma for layout, ChatGPT Images for illustrations with legible text.
Short videos: write the script with AI, but your own face and voice still work best. Dedicated tools handle captions and cuts automatically.
Generated ads and videos: Google Flow (Gemini Omni) and platforms like Higgsfield with Seedance produce cinematic scenes. Great for products and atmosphere; if a scene looks authentic and shows real people or events, it must be disclosed.

Real people and brands

Don’t generate images of real people without consent, and beware of other companies’ trademarks. For realistic images published for information or commercial purposes, add a note such as “AI-generated image”.

Ads and campaigns

AI helps write ad variants to test (headlines, copy, calls to action) and analyse campaign results exported to Excel. Ad platforms now build in generative tools too. The rule: generate many variants, but you decide which messages truly represent your brand, and measure.

What to measure

Saves and shares over likes: they signal usefulness.
Comments with questions: they become your next pieces.
Leads and enquiries, not just followers.

Each month ask AI to analyse your exported stats: “which topics and formats brought the most saves and leads? what would you stop doing?”.

In practice

Your 20 topics and voice profile, once.
A monthly plan with AI, then one strong idea a week turned into several formats.
Always reread: cut sentences anyone could have written.
Disclose deepfakes (content that looks authentic and depicts real people or events) and, by choice, other generated content too.
Measure saves and leads, not vanity numbers.

Informational content, current as of the date shown. Prices, models and rules change often: always check official sources. Written with the support of AI tools and reviewed by the author.