David Courtier-Dutton, 12 March 2025
In the next 24 months AI will impact all aspects of marketing from ideation, content creation, personalisation, data analysis and market research.
It is inevitable that this will extend to the widespread use of AI music. Indeed, a recent major report from Goldmedia predicts that, within 3 years, 27% of current music creators revenues will be replaced by generative AI.
Why AI-Generated Music will be irresistible for Brands
If, for the moment, we ignore the unfortunate economic destruction this will mean for the music industry in general, and legions of artists in particular, the rationale for using AI music in marketing is compelling.
If we assume a level playing field, ie that AI music will ultimately match the quality and appeal of human composed/performed music (remember that the AI music generated today will be the worst you will ever hear). The case for AI music in marketing is compelling:
| Existing
| AI future |
Cost | Expensive/occasionally extortionate/often priced on ‘I saw you coming’ basis
| Almost free |
Time | Clearing rights takes time
| Instantaneous, brand enjoys perpetual IP ownership |
Campaign Fit | Finding a track with correct lyrics, emotion and style can be tortuous | Using a multi modal AI model music can be composed to fit the advertising execution in terms of pace/video sync, mood, length and lyrics
|
Brand Fit | The ‘right’ track for a campaign may not also be ‘on brand’ . ie it delivers on campaign attributes but is divorced from core brand positioning | With a testing/AI feedback loop with eg OnBrand, any AI composition can be musically tweaked to land a perfect brand and campaign fit every time
|
Measurement | Subjective | Subjective control but with objective validation |
But AI music adoption will be slower than expected
Despite the compelling benefits, adoption of AI music in marketing will be comparatively slow:
· It is likely that AI will first be applied to the quick, higher value marketing wins (ideation, image and video creation, targeting etc) and music will then follow
· AI will soon enable/is already enabling the complete automation of original content creation (including the music) – particularly for high volume, low cost social campaigns. For some, to separately adopt an AI music strategy may be superfluous for all but the largest campaigns
· The perceived unresolved legal status of AI-generated music.
And it is this legal uncertainty that is perhaps creating the biggest pause. There has been widespread press coverage of the legality of AI music on the basis that the generative models that create it have been trained on music without the consent of the rights owners.
While Suno and Udio, the market leaders, are locked in a monumental case with the labels that should establish whether using copyrighted music to train models is permissible, the reality is that AI generated music is here to stay. There is no need to train models on copyrighted music, indeed Meta’s MusicGen was trained on 400,000 properly licenced tracks (approx. 0.1% of the music currently in existence). AI does not need all the world’s music to train, and quality trumps quantity every time in AI training. Perfectly good music Gen AI models can be trained on as little as 10,000 tracks – and, to put this in perspective, 100,000 new tracks a day are being uploaded to Spotify alone.
As an aside, while I have immense sympathy for (and applaud) artists fighting the UK’s proposed Copyright and Artificial Intelligence Act, whether they win or lose (or whether they all opt out) will not make a jot of difference to the rise and rise of AI music. There are plenty of publishers who are happy, for a small fee, to make their music available for AI model training. 99% of the world’s artists can slam the door but the AI horse has already bolted. However inequitable, Artists (music and otherwise) are going to be largely disenfranchised by AI. Likewise software developers, accountants, lawyers, architects, almost all computer based admin jobs and factory workers (to name but a few). And there is nothing that any of us can do to prevent this.
Even today, brands need not be concerned about using AI generated music, even if the labels can show that the model training breached copyright, it is a giant leap to prove that the original compositions created by the models somehow plagiarise the training tracks. Just like humans, the models learn how to compose music from ‘listening’ to lots of other music, they do not copy what they have ingested. On this basis every human composer is as guilty as an AI model as they too have ‘learnt’ music composition from listening to music created by others.
But is AI music second rate and devoid of emotion?
Excuse me while I hop on (one of) my hobby horses. The emotion, time and effort poured into the creation of a song is rarely reflected in the emotion and appreciation felt by the listener. Indeed most original music is mediocre regardless of the blood sweat and tears poured into its creation (and we have tested over 300,000 tracks with consumers). The fact that the artist has wrung herself out emotionally to create a song does not make it either ‘good’ or valuable to consumers (similarly, almost everyone can kick a football but a miniscule minority will ever make a living from doing so). The value of a song rests on consumer resonance and the ability to trigger a meaningful emotional response. Of course this is sometimes impacted by a personal connection/adulation of the artist, but in marketing, 99% of the time, music is chosen for its intrinsic ability to communicate emotion, not on the artist who wrote or sang it.
Last year we conducted major studies (with Stephen Arnold Music) to answer this very question ( Study 1, Study 2). The result – AI songs are often as good as, or better than human compositions, and while they are not yet that accurate at conveying the desired/instructed emotion, when blind reviewed by humans they have the ability to genuinely emotionally move the listener.
To wrap up
We are in the foothills of AI music creation and, as previously stated, the AI music you hear today is the worst you will ever hear. I have no doubt that, given a couple more years and a further dollop of VC capital, computers will soon reliably outperform human composers almost all of the time. Plug this into an iterative improvement loop with an emotional analysis OnBrand platform (or similar) and the ability of a brand to create an AI track with an entirely predictable human emotional response for commercial use will become the go to approach to music for a large percentage of campaigns.
Remember that brands do not ‘like’ music per se – they like music for the impact and emotional conditioning it can subconsciously elicit in a consumer. It is the response, not the cause that matters. When AI can demonstrably do this better, faster, cheaper and more effectively than humans, widespread adoption is inevitable.
Forrester predicts that 70% of businesses will deploy AI in marketing within the next 12 months. The Digital Marketing Institute predicting that the AI in marketing size expected to grow to $217 billion by 2034.
Kommentare