We let ChatGPT write our website - has it taken all our jobs?
Keen to embrace the future (and to understand if we’d still have jobs with all this AI) we got stuck right in to the ChatGPT.
Our use case was to improve organic traffic to our ‘CMO As A Service’ landing page and we gathered ChatGPT insights along the way that we think interesting enough to share.
Read on to discover how Drake & The Weekend, SimilarWeb, Nick Cave, The Daily Mail, hardcore SEOs, Algorithmic Watermarking, Ryan Reynolds, Prompt Engineers, and more contribute to the emerging understanding of ChatGPT. We’ll also reveal whether we still have jobs and also how our new CMO as a Service page has performed : )
1. It is breathtakingly effective
ChatGPT can blast out credible essays, copy, tables, and SEO content with incredible ease. ‘Write me an essay on the mobile phone market in [country] and write me an essay on how [brand X] should increase market share’ will produce a credible summary – there are rock solid foundations but limited detail.
2. Its growing very quickly
According to estimates from Similarweb, ChatGPT hit 100m active users in January, making it the fastest growing consumer app in history.
3. The Rise of the 'Prompt Engineer'
Last decade was about ‘learning to learn’, as we moved from ingesting/regurgitating books to getting good at searching for streams of knowledge. The next decade is about ‘learning to ask’ – where ‘Prompts’ (i.e. the detailed description of what we ask the AI for) is the new skill. We moved from asking for ‘Write me an essay on..’ to assembling a 100-word instruction Prompt for our ‘CMO services’ requirement. The difference in output was extraordinary.
4. Algorithmic Watermarking
Google and friends have 'Algorithmic Watermarking' capability – i.e. the ability of their machines to identify (and in the future maybe penalise) publishers who use other machines to make AI content. This seems to be hit-and-miss in practice, but it feels logical (and hopeful!) that human-layered quality on top of AI will always be required to drive performance in digital marketing. As for school homework, that’s a harder question!
5. Human touch always required (& Nick Cave has suffered)
Aside from the desire to not get busted by the ‘AI Police’, we obviously want the content on our site to be differentiating, very valuable, and helpful. We have lived through suffering and ChatGPT has not (as Nick Cave comments in his response to ChatGPT writing a song in his style). We have also lived through many, many decades of business experience; layering this into and on top of the ChatGPT output was essential.
6. Language Corpus to Data Corpus
V1 of the ChatGPT is apparently trained on (and we quote) ‘a corpus of human language’ – i.e. the words out there on the web. The results of that are as mentioned, very impressive – but hold tight. V2, out later this year, has been eating the world’s data. The ChatGPT that has been trained on ‘the corpus of data’ is, according to those that know more than me, going to be very impressive/scary.
7. This type of AI is an excellent way to compare long term bias between publications
Because ChatGPT is trained on years of content, the answers it gives illuminate the publishers underlying stance on its beliefs – if for example, you ask The Sun, The Times, The Guardian, The Daily Mail ‘who is Keir Starmer’ or ‘who is Meghan Markle’, the outputs are powerfully/shockingly contrasted. There could be some interesting applications for this given the right use case. (NB thanks to this insight goes to a Linkedin feed post from 14th Feb that we could not retrace to credit).
8. We love Ryan Reynolds & Mint Mobile
We work a lot in TMT and Ryan Reynolds' Mint Mobile marketing is much loved and admired here at the Bridge. Ever at the bleeding edge, Ryan asked ChatGPT to write his latest ad (all in the pursuit of saving Mint customers marketing dollars). Ryan nailed the PR, value prop, and viral marketing Trinity with this one.
9. Did it feel like cheating?
The key point in our use case – of writing web content to help people understand what CMO As A Service is, and why we are good at it, is that we know this topic very well already. We put our experience and effort into the Prompt Engineering – to define the questions and content spec we wanted as the AI output. The AI saved us hours of grinding out foundational copy, and added a layer of technical SEO consideration that would have meant another step for us to do manually. We were then able to refine, check and tune that output, before weaving our own human experience in.
If we had asked AI to write us 2000 words on Italian Jumpers, it would have been different. We know nothing about that and if we were using AI to help present us as experts in that space it would be a bit dishonest. Spotify have pulled the AI generated Drake/The Weekend track, and German artist Boris Eldagsen has rejected his (AI generated) prize winning photo in the Sony World Photography awards. These two cases highlight the rise of AI fakes among the consumer consciousness.
10. Will we still have jobs? : )
There's no doubt there is a use case for this type of application and it will reduce the time many humans spend researching and writing. But this feels like a continuation, or latest chapter in the 'Future Of Work' and 'Digital Transformation' themes. Human experience, judgement, and emotional intelligence are still vital to making the outputs high quality.
As one of our founders continuously asserts, we must embrace new technologies as they will make us 'more human' - that is allow us to think bigger, innovate, use our time more efficiently and evolve our capabilities, while 'the machines' release us from the more mundane daily tasks. Just look at the epic rise of RPA (robotic process automation) spearheaded by businesses like UiPath.
11. Was the final page any good?
Well, our CMO As A Service page is indexed by Google and has lots of organic impressions and clicks already – see what you think, and lets us know!
(And no, Chat GPT didn't write this :-))