If the problem with artificial intelligence (AI) isn’t the technology itself but the way it’s used, where does that leave marketers? In practice, there are some genuinely useful applications – and some very obvious traps. The difference usually comes down to whether AI is supporting your thinking or replacing it. Below we will cover some of the best practice guidance and general tips for using AI for social media marketing.
Where AI can help with social media marketing
- Generating alt text. Alt text is a caption that describes an image you post, useful for people with very slow connections, vision impairments, or as a backup descriptor if a page breaks and fails to show an image properly. Alt text is more or less just describing what’s in an illustration or photo, which can be a dull, time-consuming and repetitive manual task. Using AI for alt text is useful because it’s easy to verify whether or not it’s reading the contents of the image correctly.
- Rewording and/or shifting the tone of existing posts. AI can be useful for reframing a post while retaining its core sentiment. If you’re using the right tools, it’s your writing its revisions are based on, so you can easily tweak things that don’t sound true to yourself or your brand. With some tools, you can also upload best practice guidelines, a code of ethics, and a tone of voice document as part of a detailed prompt, if you want to include additional guardrails.
- Hashtag suggestions. Hashtags aren’t nearly as useful as they used to be for audience growth and content curation. However, if you want to include them, I’ve found AI can be helpful with generating suggestions relating to the content of your post and what other profiles are using.
- Cutting down word count. Since the need is to reduce rather than generate, the chance of hallucination is significantly reduced. AI works well as a ruthless editor and can help you push your post under the word limit without sacrificing salient points, providing you give it a clear brief.
- First-draft posts. Something to start with is better than nothing! If you’re struggling to articulate your thoughts or opinions on a key topic in an easy-to-digest format, it can sometimes be helpful to be presented with a starting point. This will likely need to be heavily refined to fit your voice, values, etc. but it can help to fire up those creative muscles now it has something to work from.
The risks of using AI for generating social media content
- Misinformation. Misinformation is thriving, and one of AI’s biggest problems is that when it’s wrong, it’s confidently wrong. If you’re using AI for research, or to summarise a long article, double-check it’s accurate. Every time. Even when you give it precise, accurate parameters to cite from, it can still hallucinate non-existent stats and sources. People have gotten in serious trouble over this. Depending on your role, business, and industry, the risks of not employing sufficient quality control can vary greatly. The worst-case scenario is negatively impacting your brand’s reputation, commercial interests, and opening yourself up to lawsuits if you share inaccurate or dangerous information. This is certainly the case for our industry, where accuracy in construction marketing is so important when talking about Building Regulations, Standards, and specification more generally. You can learn more about mitigating for brand risk in our crisis communications blog.
- Relying on image generation. Generated creative invites inconsistency, which over time seriously muddies your brand. This is particularly damaging if you’re generating images to represent a specific product that you’re selling; there is a high likelihood of the product changing in shape, colour and dimensions, even if you give it strict instructions not to do so. See this Instagram post from Evolving AI and Dwayne ‘The Rock’ Johnson as an example. AI generated creative will also align your brand to a closer association with slop, unless your usage of it is very careful and limited. In terms of commercial risk, there are some companies that have generated product renders, installation images, building assemblies, etc. I can’t stress enough that there should be a rigorous quality control process in place to ensure what you share accurately represents your products, brand values, and safe, compliant constructions.
- Erosion of brand voice and credibility. Generative AI tools tend to default to safe, generic phrasing and familiar structures which makes content easier to produce but harder to distinguish. Over time, this can dilute your tone of voice, make posts feel interchangeable with competitors’ (since it’s drawing on publicly available information on your niche/relevant topics), and potentially impact trust with your audiences who are increasingly sensitive to identifiably AI-generated content. If the majority of your social presence no longer sounds like a real person with judgement and intent behind it, engagement may hold steady for a while, but credibility won’t. This is a particularly big concern for businesses who believe that trust, authority, and technical expertise are essential values that their customer base chooses them for.
- Data security. This might feel unexpected since I haven’t mentioned it so far in the series, but it’s something we discussed at length when we started our internal AI Working Group here at Smith Goodfellow. It is essential that whatever we create or experiment with doesn’t leave us, or our clients, vulnerable to data security risks, especially when handling commercially sensitive information. A key consideration for us is ensuring that any AI tools we invest in offer additional data security assurances for businesses handling commercially sensitive information. For example, Microsoft states that, with the appropriate plan, data stays within your organisation’s secure workspace which helps to reduce the risk of data leaks. Your resources, AI prompts, and outputs also should never be used to train public Large Language Models (LLMs) or be accessible by competitors through AI Search. If you’re paying for any kind of AI licences you need to be certain that your chosen tool is fit for purpose, meets your business needs and is worth whatever you’re paying for it. Be especially leery of bonus AI social media features at an inflated price unless you’re sure it’ll be a gamechanger for your workflow and provide you with the assurances you need.
The importance of human oversight with AI use
Human vigilance is essential for AI regardless of what you’re using it for. You have to pay attention to what it produces and the mood of the platform you’re posting on. There’s plenty of value in the speed and soundness with which copy can be generated for surface-level needs (especially for low-risk social media posts), and it can certainly be a huge boon for time-poor marketers and business owners.
But really good copywriting, that’s precise in its style and intention, is too idiosyncratic to be effectively aped by generative writing. If you wholeheartedly believe AI can replace the work of copywriters and social media managers like-for-like, you believe your brand is generic. Leaning too hard on AI risks saturating your page with content that has already been made tired and unoriginal by the sheer volume of other prompters doing the same thing.
I can understand lots of people don’t feel confident writing without help, especially compared to AI’s speed of generation, but human flair (and even imperfection) is a fundamentally important step to build trust and a good reputation. At its worst, AI writing laminates over sprouting chances to show your brand’s distinct voice with its tepid generalities of phrasing – and AI image generation can be similarly alienating.
Content that makes people trust you and your company is content which looks like someone sat down and deliberately crafted it. This is why, if you use AI extensively, you should never use it thoughtlessly. There’s only so much human input that can be cut from a process before the end result is hollow and forgettable. In fact, I believe a characterful human-run page which uses more emojis than is perhaps necessary, slightly dodgy capitalisation, and an overuse of exclamation points, will likely perform better on socials than a trimmed, bloodless and rote AI-generated stream of posts.
As mentioned at the start of this article, AI isn’t going away. The social media slop problem will probably get worse before it gets better. But it’s still important to keep an eye on how this still-new technology is developing, even if you can’t see yourself ever using it. Ignoring a powerful technology will only hamper your ability to navigate a world that’s inarguably changed by it.
It’s still speculative what AI is all leading to. In the future, its generative capabilities might get so good that any previous errors we can point out today will be totally erased in a decade. The scale of improvement, for now, is relentless; just look at the Wikipedia article on the Will Smith Spaghetti Test for ample evidence of that.
Even if slop is eventually cleaned up, generative tools improve beyond recognition and they finally introduce some robust legislation to provide appropriate protections, the core point will still stand: if AI can replace your social content entirely, your content was never distinct enough in the first place. Human expertise, judgement and character are what make people trust you and your brand.
If you would like to explore the other posts in this AI and Social Media series, you can visit them at the links below:
Part One: How AI is Changing Social Media Marketing for Small Businesses.
Part Two: What’s Wrong With AI-Generated Social Media Content?
Part Three: Tips For Using AI for Social Media Marketing. (you are here)
Max Britton has been part of the Smith Goodfellow team since early 2025 as our Digital & Social Media Executive. You can follow him on LinkedIn or discuss any of the content in these blogs with us on our social channels: Instagram, LinkedIn, Facebook, and Threads.
