With the launch of the imaging system dall.e corporate photographers and illustrators must now be feeling a very proximate heat (the image for this blog was generated using the system). In essence you can ask the AI for an image using ordinary language, and it will “learn” from references across the internet in order to create one from scratch, an “original” free from copyright.

The copywriting equivalent isn’t far behind.

In a way it makes a lot of sense. You’ll be pushed to find a job vacancy for writers these days that doesn’t call for “content” creators rather than actual writers, and demands “SEO (search engine optimisation) friendly” copy. It would seem logical that the best thing for creating machine readable copy would be a smart machine.

The apps are already appearing.

It’s no good saying that such copy wouldn’t read particularly well. We have already seen that computers are quite capable of producing convincing human-like copy. Formulating a brief for the machine would require more detail than asking for an image, though presumably you could feed source material into the mix, give it some keywords and core ideas and off it would go.

Quite possibly too the machine could plug into data driven insights about audience segments, tailoring details to known preferences. Copy segmentation could feasibly become far more granular than is practical at the moment.

There is one big problem, for now and possibly for the foreseeable future. Google doesn’t like AI generated text, and with some irony has AI tools to detect it, fairly reliably.  

This isn’t quite the showstopper it might seem. Not all copy needs to be destined for the internet, or at least not all of needs to be subjected to the scrum of SEO. AI generated copy might quickly find a place fulfilling run of the mill text requirements, for instance in product manuals, or catalogues, or even press releases.

The AI tools could even be useful to writers in managing research, in automating the task of trawling source material and organising it as an initial narrative. The brief to the human writer would then be to translate such a text into a form that other robots would recognise as not from one of them. (That’s not necessarily as easy as it sounds.)

It’s also true that factory automation hasn’t destroyed all repetitive jobs: there are tasks where humans remain more adept than robots.  Being non-mechanical AI doesn’t have to contend with physical limitations, but I imagine there will be a version of the robot problem for copywriting tasks, where it’s simpler to brief another human than to try to turn your requirements into something a machine can work with.

Not least there’s always the business of creativity, or at least imagination. You could expect an AI to be as diligent as a human in picking out relevant material, but more imaginative writing typically makes unexpected connections. Perhaps what’s going on will prove analogous to the development of mass manufacture, at the expense of artisan craftsmen, which the likes of John Ruskin so railed against at the beginning of the nineteenth century. A couple of hundred years later, while most of what we have and consume has come out of a factory we still value artisan craft more highly than the factory equivalent.

All is not doom and gloom then. We can expect there to be a lot more AI generated commercial writing in the future, and frankly it may even be better than much of the routine communication that surrounds us. All the same there will still be a future for copywriters, even if there will probably be fewer of them. I hope they will be more valued, and better paid (I would say that).   

Predicting the future is usually perilous. The trajectory of the technology might be reasonably clear, but the ways we adopt that technology will always be subject to unforeseen or less predictable factors.

I’ve been speculating on how AI might affect of work of commercial writers in a context currently dominated by SEO , but suspect that the way social media has come to shape how we think about commissioning writing is likely to change too.

Social media is an immature phenomenon. It’s been driven substantially by the promise of a free lunch, for its general users. By and large it subsists on a promise to advertisers, that a platform will offer segmented audience targets based on sophisticated personal data gathering.

Whether social media has delivered on its promise of a continuing closer relationship with a given customer base is open to doubt. There are some clear signs that the biggest of them all, Facebook, has reached a plateau and may be in decline, particularly among younger audiences. Something else may take its place, but will have to do so in the light of increasing user concern about privacy and personal data, as well as governmental interest in platform responsibility for social harms.

These shifts are likely to play a part in undermine current assumptions about how you do SEO.

There’s an irony in the very existence of SEO. Google knows its continuing success depends on the dependability of its search results (and some transparency in showing what’s been paid for). SEO is essentially an attempt to game the system. Google’s overriding concern is to refine its algorithms so they find the most relevant, useful search results (which is why it wants to resist a deluge of AI generated copy). What a business agenda requires may or may not coincide with that concern, to put it mildly. The algorithms themselves are constantly being refined, and so we have a cottage industry of agencies offering optimisation, who in theory will be trying to keep you ahead of the changing game.

In some ways the demands of writing for SEO are not a million miles from the way a well-written news story has long been supposed to function. The opening of the story should signal immediately what it’s about. The hooks designed to get people to read on should be front loaded, with explanatory or amplifying detail in what follows. SEO priorities look similar, though there will be additional pressure to get stuff in about the product and brand. The current limits of the algorithms bring other distortions. Have you ever searched for a recipe and wondered why you have to wade through paragraph after paragraph about the recipe writer’s life in order to get to the food? Yes, that’s SEO at work, because the algorithm favours longer copy, assuming it will offer greater depth, rather than extended waffle that nobody really wants to read.

For Google’s purposes, the algorithms should be moving towards a state where it becomes harder and harder to game the system, where there is no short cut to relevance: if you want to get noticed you need to offer copy that people might actually want to read. Promotional messages become an unwelcome distraction, or at least the message becomes subliminal: that the brand behind the material is adding value to a relationship, taking that relationship beyond the merely transactional.

In a way this has long been the ambition of brand builders. The whole point of brand building after all is a to build a customer allegiance that will reduce price sensitivity. That’s fair enough, but it takes commitment and investment, and will always be harder to sustain if margins are under the pressure of a depressed economy.

Social media on the face of it offer a ready tool for sustained audience engagement: for a potent combination of informative or entertaining material, and a platform for interaction. But these platforms need to find a way forward both in terms of editorial standards (however much they are resisting the notion that they are publishers, with a responsibility to truth and social cohesion, a resistance doomed by the fact that it flies in the face of reality) and in their funding model. The promise of micro targeting for advertisers is foundering on the twin rocks of rightly increasing privacy concerns and advertiser scepticism about the promise itself. Again the future is very unclear here, though I’d guess it will depend on a shift to anonymised data (realistically promising only a broader but still useful audience segmentation).

I’d also imagine that we’ll see more and more publication sites taking a stand on the need for readers to pay for valued content, or at least offer a choice between having to watch ads and paying a sub.

I hope and believe that funding models for internet sites will have to learn to do without micro targeting. This will if anything increase the demand for SEO, but further improvements in Google’s search algorithms will change its nature, increasingly emphasising the importance of relevance. It will be fascinating too, to see whether Google can develop automated ways of assessing good quality: if it can distinguish between AI and human generated text perhaps this won’t be such a stretch.

There’s always comfort in the mechanical, in the idea that if you press the right buttons you’ll get a predictable response. But I don’t believe communication, or even language, can work this way. Humans are too complex, the contexts in which communication must work too varied. You can narrow the scope for misunderstanding, but I’d argue that good writing has to go further. Good writers need more than craft skills: they need the intuition and creativity to imagine how an audience might feel, and then harness the power of language to evoke those feelings. As things stand I don’t believe you can program that ability into an AI, not least because it depends on consciousness. It’s a live argument right now as to whether an AI could develop consciousness, but it’s an argument that always seems likely to struggle with the reality that we can’t really say what consciousness is.  

There’s a conceptual problem with AI writing, a mistaken assumption that creating effective writing is a matter of assembling the correct content and organising it coherently. I’ve agreed that there are contexts where this will be adequate, and perhaps an improvement on some current practice, but there’s more to good writing, and that’s why humans will continue to have a role for the foreseeable future. More than this, good imaginative writing can help to create an audience as well as follow it. If the real goal is a sustained relationship, then you need that human power in your voice.

Oh, and this piece has not been written by an AI, though an AI probably would say that. Deceitful buggers, all of them.