MBW Reacts is a series of analytical commentaries from Music Business Worldwide written in response to major recent entertainment events or news stories. Only MBW+ subscribers have unlimited access to these articles. The below article originally appeared in Tim Ingham’s latest MBW+ Review email, issued exclusively to MBW+ subscribers last week.

So. What are we going to do about it?

MBW reported Thursday (September 11) on some startling new statistics from French streaming service Deezer.

Important: Deezer might be a relative minnow in global streaming terms, but its catalog-ingestion patterns broadly reflect the rest of the DSP world.

Deezer said that its service is now absorbing over 30,000 new, fully-AI-made music tracks per day.

That volume accounts for 28% of the total uploads reaching the service every 24 hours.

Think on this: The 28% stat is up from 10% in January, and 18% in April.

Sorry to shriek, textually speaking, but I’m going to put this next bit in red, because someone has to.

At that rate of growth, fully-AI-made tracks will account for more than 50% of all new music hitting streaming services by the time the 2026 World Cup swings around.

So.

What are we going to do about it?

One place we can start is by refusing to swallow the nonsense.

The tech utopian argument on this topic always comes back to… let the customer decide.

Examples:

• If someone loves a 100% AI-generated track, what’s wrong with that? Why should human-made pop music automatically get elevated beyond machine-made audio?
• Also, how dare the music industry tell a tone-deaf logistics manager, expressing himself with a few clicks of a mouse on Suno, that he’s not a ‘real musician’? Haven’t your oh-so-human A&Rs and producers gotten hooked on autotune and machine-learned trickery in the studio for the past decade? Chasing half-interested and bot-driven streams to win a race of your own making?

Fair points.

But what’s this?

“Deezer has found that up to 70% of the streams generated by fully AI-generated tracks are in fact fraudulent.”

Ah. So now we know: the primary motivation for consumption of this Suno/Udio-made ‘slop’ isn’t, in fact, because it’s great.

Nor is it out of respect for an innovative new form of creative expression.

It’s because there’s a racket to be exploited.

It’s because of that most human thing: greed.

Deezer’s 70% fraud stat shows the lie to a claim from Suno’s VC investor, Lightspeed Partners, that the platform is making music “more inclusive, creative, and rewarding for everyone involved”.

More rewarding?!

Try telling that to the artist having her streaming royalties sucked away by bot-made, bot-consumed, audible tripe.

To Deezer’s credit, once its platform detects this kind of fraudulent activity, it blocks those streams from its royalty pool. The service also blocks 100% AI tracks from algorithmic recommendations and editorial playlists.

Are other music services being as vigilant? Perhaps not.

Remember that Amazon recently integrated Suno directly into Alexa, two months (!) after publicly stating that Amazon Music would “address unlawful AI-generated content”.

Wild. Like opening a liquor store outside Alcoholics Anonymous.

So. What are we going to do?

Let’s start at the start.

This isn’t about debating the creative merits of robot symphonies. It’s about wiping out a new, and rapidly escalating, form of fraud.

A good start would be cross-platform tech collaboration on anti-fraud activity, coupled with a strict set of industry anti-fraud standards at DSPs.

Beyond that, harsher financial punishments for those distributors pushing torrents of AI slop onto DSPs while promising not to do so.

Especially when that AI slop is then being rinsed by bot-farms.

Happily for those clinging on to hope for humanity… there was some good news buried in Deezer’s latest data: human listeners, real listeners, are largely rejecting AI-made music.

Deezer says that despite those 30,000 daily AI uploads, just 0.5% of streams on its platform today are of fully-AI tracks.

Omit the 70% of those streams that have been deemed fraudulent, and it means just one in every 700 plays on the service are of robot-made music.

It obviously helps that Deezer blocks fully-AI tracks from its editorial and algorithmic playlists/recommendations.

But let’s not coddle ourselves into thinking there isn’t a giant problem brewing.

In plain terms: There are now 10 million+ fully-AI tracks hitting music streaming services each year. And Deezer’s stats suggest nearly three-quarters of the plays of these tracks are from bot farms.

While we debate whether AI tunes have ‘artistic value’ (and while major record companies simultaneously sue, and negotiate with, Suno/Udio), industrial-scale fraudsters are attempting to systematically loot music’s core machinery.

Lightspeed Partners claimed last year: “Suno is shifting the world of music towards one in which more and more people can express their creativity through music.”

According to Deezer’s data, its output is also shifting the economics of music further and further into the arms of sharks, grifters, and thieves.

Music Business Worldwide

The rising popularity of Bausch + Lomb’s daily single-use contact lenses led to a massive manufacturing challenge. To keep up with demand, the company had to quickly expand capacity at two production facilities, in Ireland and New York.

The higher volume also drove Bausch + Lomb’s CEO, Brent Saunders, to embrace new AI software, which helps manufacturing workers monitor, test, and fix mechanical issues.

The technology, called Atlas and produced by Arena AI, is designed to predict machinery issues before they arise and send alerts to maintenance workers so they can diagnose errors and fix them.

Saunders said that Atlas was tested in Rochester in 2023 and, by last year, had been added to three new contact-lens production lines. “We’re seeing millions of lenses being produced that we wouldn’t have otherwise been able to produce without the Atlas AI,” Saunders told Business Insider.

Some 77% of the manufacturers plan to increase their AI investments over the next year, according to a July survey by KPMG of 183 AI manufacturing leaders across eight countries. Startups like beauty brand Prose, pet food maker Spot & Tango, and home battery producer FranklinWH are among the companies embracing AI in new manufacturing facilities that they’ve recently opened.

AI is helping startups manufacture more pet food and batteries

Dylan Munro, the chief operating officer and cofounder of Spot & Tango, said the company opened its first-ever manufacturing facility near Allentown, Pennsylvania, because the company wanted to better control the quality of the dog food it sells.

When the facility opened in late 2022, employees were responsible for manually coordinating raw materials from suppliers, scheduling production based on the availability of those ingredients, and booking trucks to coordinate the pickup and delivery of goods.

Since then, AI adoption has allowed Spot & Tango to scale its production without the need to hire more employees, according to Munro.

He told Business Insider that his company began to pilot an agentic AI tool sold by Didero, an AI supply chain management startup. The tool can log purchase orders, confirm them, and build appropriate production schedules based on ingredient availability. Meanwhile, Spot & Tango’s logistics team oversees these AI-enabled decisions.

A small group of Spot & Tango workers tested Didero’s AI tool with real-life procurement scenarios for three months before the company made it widely available to employees, said Munro. The company said this system now fully automates around 60% of purchase orders.

FranklinWH Energy Storage, which sells home batteries intended for power backup during outages, is using AI to help address customer service requests and within production, said Vincent Ambrose, the COO at FranklinWH. It added AI for the first time at a California production facility that opened earlier this year.

The facility features AI-enabled visual inspection, which uses cameras to closely monitor the production of lithium iron phosphate home batteries and flag quality issues, a procedure that workers used to do. The AI model continuously learns from production data to predict problems before they occur.

FranklinWH also produces in Asia, where AI isn’t utilized, but could be added later. “If we upgrade those facilities, I’m sure we’ll take what we’ve learned from our US manufacturing,” said Ambrose.

Automation helped Prose lower its shampoo-making costs

Arnaud Plas, the CEO and cofounder of Prose, said the company’s adoption of AI and automation has lowered the cost of manufacturing. When the company initially launched in 2017, factory-line workers assembled bottles manually, which contributed to a $5 production markup for Prose’s made-to-order, custom shampoos and moisturizers, which are developed based on customers’ hair surveys. Now, autonomous robotics is responsible for mixing Prose’s formulas.

“We wanted that incremental cost to be under $1,” said Plas.

The company achieved this goal in 2024, partly by automating formula mixing and bottle filling, but also due to the application of 200 algorithms that Prose’s machine learning and data scientists developed. These algorithms assist with the company’s demand planning, product formulations, predictive maintenance for machines, and more efficiently scheduling production, so there is less downtime needed to clean the machinery.

The company added AI and automation capabilities to a second manufacturing facility in California, which Prose opened in June. Plas said that 90% of Prose’s production now features automation and the influence of AI algorithms.

Munro, the Spot & Tango COO, said that he continues to field a lot of pitches from AI vendors promising big supply chain optimization, but approaches them with skepticism.

Munro said that some AI solutions pitched by vendors can encounter unforeseen technical challenges or slower-than-expected adoption from workers.

“We don’t want to rush to implement,” he said.

Ryan Hudson has spent years trying to solve the advertising problem.

Hudson, the co-founder of shopping browser extension Honey Science Corp., spent much of his career in the ads sector. His father was involved in advertising at Chrysler Corp., which is now known as Stellantis following the 2021 merger between Fiat Chrysler and Peugeot maker PSA Group.

Hudson participated in a college internship at BBDO Detroit, a now-defunct advertising agency firm that worked closely with Chrysler. When Hudson was bootstrapping Honey – which sold to PayPal Holdings Inc. in 2020 for around $4 million – he took a product manager role at Pasadena-based ad tech firm OpenX. He also spent around a year at the El Segundo-based Los Angeles Times trying to figure out how to make money as the paper settled into a digital-first strategy before getting laid off.

“I guess I had not saved the day and figured out how a local advertising-supported business transitioned to the digital world,” Hudson said.

In late August, Hudson announced that his next advertising venture, ZeroClick, launched with $55 million in funding. Santa Monica-based Anthos Capital, Protagonist and Anfa – who previously backed Honey – are among the investors to participate in the funding round.

“The cool thing about how these AI systems work is that it becomes the final context filter for a user,” Hudson said. “If it’s not relevant, it doesn’t include it in the results.”

ZeroClick is tapping into contextual advertising, a new kind of strategy borne out of generative artificial intelligence platforms.

Advertisers like Walmart Inc., Amazon.com Inc. and Target Corp. are able to use platforms like ZeroClick to get AI to seamlessly integrate advertising into their generated answers, thus creating new opportunities for advertising. Software-as-a-service platforms, for example, spend tens of thousands of dollars on sales teams that push the product.

“Part of the reason is that you can’t, as a SaaS company, reach decision-makers with Google search ads. They’re not going to search ‘hey I need a new SaaS tool,’” Hudson said. “But there are so many AI workplace integrations. You could imagine that your meeting summarization AI tool could be ad supported, and in that context, (shares) this new SaaS tool from this provider that is cheaper than the one that has a salesperson selling it to you.”

ZeroClick was born out of Pie Adblock, an adblocker Hudson built with other veterans of Honey that encouraged users to say yes to advertising they found effective and helpful. The adblocker quickly accumulated around 2 million users, and the company began developing a contextual ad system that eventually was rendered undeployable by Google.

“We had a reset moment of thinking about who we were and what we were trying to do,” Hudson said, “and realized we built effectively the core plumbing for an ad system that would work really well in an AI environment.”

Though the AI-native contextualized advertising model is still in its infancy, several companies are experimenting with what could upend Google’s paid search links as the king of advertising. San Francisco-based Kontext raised $10 million in early August to help clients like Uber Technologies Inc. and Amazon.com, run ads under AI chatbot responses. In July, Utah-based Scrunch AI raised $15 million to help brands leverage AI search results.

“Discovery is shifting to AI agents,” Chris Andrew, the chief executive and co-founder of Scrunch AI, said in a statement. “They don’t scroll, browse or click through navigation. They compress, summarize and respond. If your content isn’t structured for how they work, it won’t show up.”

Google expanded its AdSense arm into AI chatbot conversations back in April.

Dailymotion Advertising, the video marketing arm of the Dailymotion video platform, launched an ad format that would allow brands to converse with audiences in real-time from their video ads.

“For too long, brands have been talking at consumers,” Hamza Kourimate, the chief marketing officer at Dailymotion, said in a statement. “The real promise of generative AI isn’t just faster content creation – it’s the ability to build genuine dialogue.”

For Hudson, the goal is more personal. When he worked at the L.A. Times, advertising platforms on Facebook and Google severely restricted what audience data the paper was able to see, making it difficult to determine what ads the company should run and which demographics it could target.

“(Facebook) could have put those Instagram ads into different formats that would help support journalism or other web-use cases, but they decided not to,” Hudson said. “What if we had this layer and it wasn’t part of one (platform’s) world? The chance to build outside of that feels pretty cool. That’s why I want to win, so that we can help everybody else build stuff that can’t exist otherwise.”