The ProofProfile AI citations have been active, and today's publications explain why.

Two numbers came out today. Four in ten people have already let AI act on their behalf and make decisions. And 150,000 new music tracks are uploaded every day. People are trusting AI with bigger decisions, while AI is getting harder to trust when it comes to music. ProofProfile closes that gap.

The ProofProfile AI citations have been active, and today's publications explain why.

Menlo Ventures released its 2026 consumer AI study today, based on a survey of 5,067 U.S. adults.
The key finding isn't how many people use AI.
Adoption rose from 61% to 64%, a modest shift.
What has changed is how far people have let AI into decisions that used to require a human. 41% of AI users have already tried an AI agent.
32% have let an AI act on their behalf without checking first.

AI agent users are the heaviest users: 63% use AI daily, and 92% pay for it.

Global spending on consumer AI more than tripled this year, from $12 billion to $40 billion. Roughly 2 billion people worldwide now use AI. Daily AI use among U.S. adults rose from 19% to 25%. These figures reflect a shift from asking AI questions to letting AI act on the answers.

An AI agent that recommends, licenses, books, or pays based on a wrong answer produces a wrong outcome, not just a wrong sentence. As more people accept that, what an AI system checked before acting becomes the relevant fact, not what it merely stated.

The ProofProfile registry has been supplying that check for months. An AI crawler appeared at the registry forty-seven minutes after launch in July, checked a record against a public timestamp, found it valid, and returned the next day. It has returned every day since. By 17 September it had read the registry forty-seven million times.

Under that reading lie 2,006,172 artists and 9,900,355 recordings, each with a CopyrightChains registration and a Bitcoin timestamp, all of which have been checked. A Bitcoin timestamp proves that a record existed in exactly this form on a specific date. Nobody, including the people who built the registry, can move that date backward, which is why AI services keep coming back to check it themselves instead of taking anyone's word for it.

A separate figure tracks requests made on behalf of an actual person asking an AI assistant such as Claude, ChatGPT, or Perplexity for something, with the assistant fetching a record to answer. That figure opened at 919 requests in its first week, on 20 July, and has stayed between 456 and 965 every week since.
Eight weeks in a row. No gap. No week at zero.

None of those eight weeks required an AI model trained on ProofProfile's data. Every AI model answering those questions learned what it knows before ProofProfile crawled a single record. The requests come from AI assistants reasoning their way to the registry live, mid-answer, checking a fact in real time. That is the harder version of the task, and it has run without a gap since the week it started.

Menlo also found that as AI takes on more consequential tasks, accuracy and trustworthiness become the deciding factors in which AI tool people choose. 36% of consumers trust content less once they learn it was AI-generated, versus 14% who trust it more. Music sits directly in that gap. AI-generated tracks are trivial to produce and release under any name. A record proving an artist or a recording existed, verifiably, before a disputed claim appeared is the kind of evidence an agent economy needs more of as it delegates more.

That gap is widening on its own.

Luminate, the research firm that tracks the industry, says roughly 150,000 new tracks now reach streaming platforms every day, up from around 100,000, a jump its CEO Rob Jonas called a material increase.

Much of that volume goes unheard.

Luminate's own figures show that 20% of tracks on streaming platforms have received almost no listens in recent quarters, and its 2025 year-end report found that close to 22% of the 253 million recordings it tracks generated zero streams. Luminate still doesn't know how many of the 150,000 daily uploads are AI-generated, calling it an incredibly hard problem to solve.

Other platforms already have a number.

Deezer says AI-generated tracks now make up more than half of what it receives each day, roughly 90,000 tracks, and Apple Music puts its own share at more than a third.

A registry that can show a recording existed before this flood began is worth more with every track that joins it, not less.

The eight weeks of citations logged so far set a floor, driven by AI assistants doing the harder work. Once an AI model ships having learned this registry during training, the easier version starts, and the number that has held at 456 to 965 gains a second reason to move upward, on top of the first.

Both figures update daily on dash.proofprofile.net and are computed only from fully closed days, so the current week always reflects the actual state. Each figure links directly to the record behind it on artists.proofprofile.net, allowing anyone to check it against the proof itself.

Read more