At a glance: When buyers can't tell good from average, they stop paying for good and start paying less.
Two researchers at UCLA tracked 49,000 freelancers across five years and found that after ChatGPT arrived, the credentials, the portfolio and the track record started counting for less, and price started counting for more. That's commoditisation measured rather than feared, and it's bounded: your expertise is worth less in the one place where a client meets you as a profile and a price, with no relationship to go on. Your rate has stopped doing the signalling work it used to do, so cutting it won't help. AI wants to make you interchangeable — here is the argument in full. Meanwhile the same technology is shrinking the pipeline that produces people like you, which makes what you've built rarer at exactly the moment it's getting harder to prove. Pricing power is downstream of positioning. Move the work out of the channel that flattens you, and the erosion never reaches you. |
The reward for being demonstrably better did not vanish. It moved. It left the places where buyers compare profiles and went to the places where they trust a person.
What the study actually measured
Auyon Siddiq and Niuniu Zhang built a record of 49,610 workers who completed 2.26 million contracts on Upwork between January 2021 and March 2026, a window that brackets the release of ChatGPT in November 2022. Rather than hand-coding what was in each worker's profile, they represented the full text of every profile as a dense numerical fingerprint, which let them measure how much a worker's accumulated background, as opposed to their hourly rate, explained who actually got hired.
Then they did something careful. They sorted job categories by how exposed they were to AI and compared what happened to hiring in high-exposure categories against low-exposure ones after ChatGPT landed. If AI were changing how the market values experience, the change should show up more sharply in the categories AI can touch. It did. In the most AI-exposed categories, the weight the market placed on a worker's human background, the certifications, the work history, the skills, the ratings, fell by about 7.8 percent relative to an unexposed category. Over the same period, the weight placed on price rose by about 1.1 percent.
Those two numbers are the whole story: the market moved a little of what it was paying for background across to what it was paying for cheapness. The effects grew over time and were largest at the end of the study, which tells you the market hasn't finished adjusting.
The headline number
In the job categories most exposed to AI, the value the market placed on a worker's accumulated background fell by roughly 7.8 percent after ChatGPT, while the value placed on price rose by about 1.1 percent. The gap is still widening.
A separate study of the same platform, reported by Aki Ito, puts a sharper edge on it. Among freelancers offering image-based services, the top earners took 7 percent fewer jobs and watched their earnings fall by 14 percent. The best were hit hardest. So the answer here can't be "be better", because being better is what got measured, and being better is what stopped paying.
This isn't only a platform story. Laetitia Vitaud, on the Wisepreneurs podcast, described the anxiety underneath it in a way the data now confirms.
The trend may be that all work could be paid less in the future, not necessarily disappear, but be paid less. And that I think is the anxiety ... Are we going to be valued enough to be paid enough to have a good life?
Commoditisation rarely arrives as work vanishing. It arrives as the same work paying less.
Why your price stopped signalling quality
There's a second result that matters more than the headline, because it overturns something experienced professionals have leaned on for years.
For a long time, a high price did useful work on its own. It signalled quality. A client who couldn't fully judge your ability could at least reason that someone charging a premium probably had the goods to back it. Siddiq and Zhang tested whether that still holds, and found it running in reverse. In AI-exposed categories, demand shifted toward lower-priced workers, not higher-priced ones. If clients were reading high prices as a mark of quality, hiring should have moved up the price ladder. It moved down.
Call it the price-signal inversion. A high rate has stopped reading as evidence that you're good and started reading as a liability, because the buyer can no longer see the difference the rate is meant to mark. On a platform where AI has compressed the visible gap between a strong worker and an average one, charging more no longer reassures anybody. It just makes you the expensive option next to a cheaper profile that looks, on screen, much the same.
Why AI tools disagree about your practice, and what that tells you about your positioning.
The instinct at this point is to cut your rate.
Resist it, because it reads a channel problem as a pricing problem. The reason your price stopped signalling quality is not that the price is wrong. It's that you're being met somewhere the buyer can't see your quality, so nothing about your number can fix it.
Cutting also has a nasty property that Robert Vlach described precisely when he came on the podcast to talk about pricing. Overcharging tells you quickly.
"If you overcharge your clients, you would soon find out because the deals won't happen," he said. Underpricing tells you nothing. "If you ask a price that is too low, you wouldn't get many negative or red flag signals, because everybody would be pretty eager to order more."
He's blunt about why, because he hires freelancers himself.
"As a client, I usually don't tell my supplier that they are too cheap. I'm happily buying at the low price."
So the discount you take this quarter produces exactly one piece of feedback: more work at the lower number. It feels like validation. It isn't.
This is also where the difference between a knowledge half-life and a wisdom half-life starts to bite. What you can look up, state, or hand over has a knowledge half-life, and AI is shortening it as the machine gets better at looking up, stating and handing over.
Knowing which of several plausible options to back for this client, this year, because you've watched similar calls play out and seen how they landed, has a wisdom half-life, and it doesn't decay the same way.
Pricing power is downstream of positioning, and positioning is about where and how the buyer meets you, not what you charge once they do. It's worth understanding why experience literally changes how you think before you ever discount it.
The signal is fading just as the substance gets scarcer
Here's the part the study doesn't cover, and it changes what the finding means for you.
While AI compresses the visible difference between a strong professional and an average one, it's also eating the work that used to manufacture strong professionals in the first place. The junior tasks, the first drafts, the grunt research, the routine analysis: those were never only production. They were the apprenticeship. Automate them and you get a short-term productivity gain and a long-term supply problem, because the pathway that turned capable beginners into people with real depth has been removed.
Jon Younger has watched this from inside the talent world for decades, and when he came back on the podcast he put it in career-stage terms. In the early independent-contributor stage, he said,
"the half-life of that contribution is only so long. And we know in AI it's even shorter than ever."
Then he named the squeeze directly.
In stage two, we were expected to know a lot, but we weren't expected to have the experience of 20 years. And that's the AI problem and that it shortens it.
Jon Younger, Wisepreneurs Podcast, episode 79
So the market is getting worse at recognising accumulated depth at the same moment that accumulated depth is getting harder to acquire.
That puts the work squarely on where you're seen, which is why the rest of this comes down to channels rather than rates.
Why this is the strongest argument for the way you already work
It would be easy to read all this as bad news for anyone who has spent decades building genuine capability. Look closely at where the study was run. Upwork is a particular kind of market. Clients contract for discrete jobs from a distance, usually without ever having spoken to the worker, choosing from profiles and prices.
The researchers say so directly: they observe the profile and the rate, and they can't see relationships, reputation built outside the platform, or anything that happens before the click. The commoditisation they measured lives inside that constraint. It's what happens to expertise when a buyer can only meet it as text and a number.
That's precisely the situation experienced professionals building independent practice are trying to avoid. When a client comes to you through a referral, a conversation, a body of work they've followed, you're not a profile being compared against three cheaper ones. You're a known quantity, chosen before price enters the conversation.
So the study isn't a threat to that model. It's the evidence for it. Upwork is the control group. It shows what happens to professionals who let themselves be matched at a distance on signals a machine can flatten.
I have built my own practice this way for years. Most of my clients arrive through a referral, or because they've read something I've written, not through a profile and a rate, and that's exactly the part of the market the study can't see.
I'm also on the other side of it. I've hired through Upwork for a long time, and I'm careful about who I choose, because what I'm buying is someone who can tell me when the brief itself is wrong, not someone who's available and cheap. In marketing speak, what a referral does is create awareness and then a preference, before anyone has quoted anything.
One honest qualification, because the researchers raise it themselves. They note the same effect could show up in traditional hiring, which also screens on credentials and references. Nobody has yet measured what's happening in relationship-rich channels. Treat relationship as the best defence currently evidenced, and keep working at it, rather than a wall that holds by itself.
Ben Legg, who runs The Portfolio Collective, a global community of 16,000 portfolio professionals, has watched the same split open from the market side.
"For lower level freelancers doing more inexperienced work, the market's been flatlining for the last five years," he told the podcast. "For higher level work, the freelancers that earn over a hundred thousand a year, it's doubled in the last five years."
The bottom of the market, the part that meets clients as a profile and a price, is being commoditised. The top, the part that works through relationship and reputation, is growing.
Why the relational channel cannot be flattened
The reason a relationship resists commoditisation isn't sentimental. It's structural. On a platform, the client hands you a specification and you hand back a deliverable, and the thing being bought already exists in principle before you're chosen, which is exactly what lets a buyer line you up against three cheaper profiles that appear to offer the same thing.
A relationship runs the other way. The value is made inside the conversation, together. You don't hand over a pre-existing thing so much as co-create the understanding of what the problem actually is, and then what to do about it. There's nothing to compare on a profile, because the work doesn't exist until you and the client are in the room making it.
That's also what the client is paying for. Information is cheap now and getting cheaper. Understanding is expensive.
What a client buys from you is a reduction in the cost of understanding their own situation, turning what they already have in front of them into something they can act on with confidence. No profile can encode that, because it doesn't exist until it meets their specific mess.
This is why the part of your expertise that matters most is the part that never fits in text. A machine reads and writes in explicit language.
The load-bearing communication in expert work happens in a register underneath that, the read of a room, the sense that a brief is wrong before the numbers say so, the reframe that arrives before you've consciously worked out why.
That difficulty you have explaining exactly how you knew isn't a weakness in your positioning. It's the evidence that what you do lives in a language beyond the machine, and therefore beyond the reach of the channel that commoditises. It's the paradox of expertise: the deeper the reading, the harder it is to state, and the harder it is to state, the less a platform can encode, rank and price it away.
Vitaud puts the human side of this plainly. Talking about what selling actually is when it's done well, she said it
"is about caring for that person, listening to their needs, understanding what they will need, and creating trust and developing a relationship with a lot of trust."
That's not a soft skill bolted onto the expertise. In a market flattening everything it can encode, it is the expertise, because it's the part no profile can hold and no model can reproduce.
Valentine Gatard on human agency and why career independence matters more as AI advances makes the same argument on the career side.
What the market still pays a premium for
The capability that survives commoditisation is the part a buyer can't assess from a profile, because it only shows up in contact with a real situation. It's reading a client's problem as a whole and naming what's actually going on, rather than answering the question they came in with.
Jean Boulton calls that mode of attention sensemaking, and it's the opposite of breaking a situation into parts and measuring them in isolation, which is the thing machines are now very good at. This is the verification bottleneck that makes experienced professionals more valuable now, not less: when AI has collapsed the cost of producing options, the scarce thing becomes knowing which option is worth acting on.
Johanna Rothman, thirty years an independent consultant, draws the line that matters here. A great deal of what gets called consulting, she says, is really contracting. Those people
"are not responsible for the client's growth at all. They are an extra pair of hands," and the work is "almost always time and materials, which is almost a useless way to really think about consulting."
The contractor is the commoditised channel in person, priced by the hour and lined up against cheaper hands. The consultant is paid on a different basis entirely.
A retainer means you get paid for not specifically your time, but for the wisdom and the strategy and the tactics you have all learned.
Johanna Rothman, Wisepreneurs Podcast, episode 92
That's pricing power downstream of positioning, said by someone who's lived it. The moment you're paid for what you work out rather than the hours you were present for, there's no per-unit rate left for a buyer to argue down.
The practical use of AI follows from this. As Ben Legg puts it, the move is to work "on the machine, not in the machine": let AI do the commoditisable production, the part that was going to be flattened anyway, so your hours go to the co-created, in-the-room work that no profile can hold.
That's not resisting AI. It's using it to clear space for the part of your practice that appreciates. One caution, which is the apprenticeship problem in miniature: before you hand a task over for good, ask whether doing it is part of what keeps you sharp at the thing you can't hand over.
Next steps
Start by working out how much of your own pipeline currently runs through commoditised channels, where a prospect meets you as one option among several, judged largely on price. That share is your exposure, and the aim over the next year is to lower it.
Here's a quick way to see where your own work sits. Take your last ten or twenty pieces of work, mark which channel each one came through, and see how much of it lands in the high-exposure rows.
Channel | How the client meets you | Exposure | The move |
|---|---|---|---|
Freelance platforms (Upwork, Fiverr) | A profile and a price, ranked against others | High | Use for discovery only, then take the relationship off the platform |
Job boards and open tenders | One name on a shortlist, judged on rate | High | Stop chasing, it is the worst use of your time |
Marketplaces and directories | A listing among many | High | Treat as a shopfront, not a pipeline |
Cold ads and paid lead generation | A rate compared with alternatives | Medium to high | Warm the prospect with your writing before price comes up |
Referrals from clients and peers | Introduced, and already half-trusted | Low | Ask for them, and make them easy to give |
Your writing, podcast and talks | They know your thinking before they call | Low | Publish consistently, this is your positioning |
Repeat clients and retainers | An existing relationship | Low | Deepen it, and move hourly clients onto retainers |
Your network and community | A known person, not a profile | Low | Invest steadily, it compounds |
The high-exposure rows are where the study's erosion reaches you, and the low-exposure rows are where your experience still commands a premium. The year's work is to shift the balance down the table.
Then look at how your week is actually allocated, because the channel question has a work-design twin. Four things compete for your hours: the work only you can do, the work that needs a human but not you, the work a machine can handle, and the relationship-building that bills nothing directly but sustains all of it.
Most of us overinvest in the middle two, because they feel productive. The first is what clients pay a premium for. The last is what keeps you out of the channel where premiums disappear.
Then invest in the channel the research says is appreciating. Legg is blunt about how independent work is actually sourced.
"Only 20 percent of fractional work ever gets listed," he said, with "a hundred to two hundred applicants for every post." Roughly half of opportunities, in his account, "come from networking, word of mouth referrals," and the goal is to "get to a point where you never need to apply for anything."
Vitaud makes the same point from the other side. The network, she argues, is now
"the most important asset to start a business. How many people do you know? How varied is the network that you have? ... That's the main asset to start a business today, because there's less capital involved in most activities."
Those relationships are slow and unmeasurable, which is exactly why most independent professionals underinvest in them, and exactly why they stay scarce and valuable. They're also what makes you findable in the way that now matters. The same relationships, published thinking and consistent presence that get you chosen before price is discussed are what get you found on Google and cited by AI when a prospect goes looking.
The professionals who will do well as AI spreads are not the ones with the most polished profiles. They're the ones whose clients never needed to compare profiles in the first place.
Frequently asked questions
Why would charging more ever hurt me, and what is the price-signal inversion?
The price-signal inversion is what happens when a high rate stops reading as evidence of quality and starts reading as a liability, because the buyer can no longer see the difference the rate is meant to mark. On a platform where AI has narrowed the visible gap between a strong worker and an average one, a premium just marks you as the expensive choice next to a cheaper profile that looks similar on screen. Cutting the rate is worse, because underpricing generates almost no corrective feedback: clients happy with a bargain rarely tell you that you're too cheap.
What does "pricing power is downstream of positioning" mean?
It means your ability to hold a premium rate is decided before the rate is ever quoted, by where and how the client meets you. Met as a profile among many, no price works, because the buyer can't see your quality. Met through relationship and reputation, the same rate reads as the cost of being right. You fix pricing by changing the channel, not the number.
Is relationship-based practice permanent protection against the AI commoditisation trap?
No, and the researchers say so themselves. They note the same effect could appear in traditional hiring, which also screens on credentials and references. What has actually been measured is the erosion in markets where buyers meet suppliers as a profile and a price. Relationship is the strongest defence currently evidenced, and it should be treated as an advantage to keep working at rather than a wall that holds by itself.
If AI is removing entry-level work, doesn't that mean fewer experts later?
That's the expectation, and it's why this matters beyond your own rate. The junior tasks being automated are the same tasks that historically turned capable beginners into people with real depth. As that pathway narrows, the kind of accumulated capability you already have becomes scarcer, at the same time as the market signal that advertises it becomes weaker. Scarcity and invisibility arriving together is exactly why positioning does more work now than pricing.
What is the difference between a consultant and a contractor here?
A contractor is hired as an extra pair of hands, judged on the deliverable and usually paid time and materials, which is the priced-by-the-hour, comparable work most exposed to commoditisation. A consultant is paid for working out what should happen, often on a retainer, where there's no per-unit rate for a buyer to compare and drive down. The same person can be either, depending on how the work is positioned.
Why can't AI commoditise relationship-based work?
Because that work is co-created rather than handed over. On a platform the deliverable exists in principle before you're chosen, so it can be compared and priced down. In a relationship the value is made inside the conversation, and what drives it lives in a register that never fully fits in text, which is exactly what a machine and a profile can't reproduce.
References
Johanna Rothman Independent Consulting: Principles That Outlast Every Trend.
Wisepreneurs Project podcast, Episode 92, on the consultant versus contractor distinction, retainers versus time and materials, and value that AI cannot reproduce.
Laetitia Vitaud Career Reinvention After 60: Why Your 150 Relationships Beat Startup Capital.
Wisepreneurs Project podcast, Episode 71, on the future of work, valuing what is human, and why relationships and critical judgment hold their worth as AI spreads.
Ben Legg Portfolio Careers and Expertise Monetisation: From Google COO to Solopreneur.
Wisepreneurs Project podcast, Episode 87, on the split between commoditised entry-level freelance work and growing demand for experienced practitioners, and working on the machine rather than in it.
Jon Younger Professional Relevance: Positioning Expertise When AI Does the Routine Work.
Wisepreneurs Project podcast, Episode 79, on the shortening half-life of technical contribution and the squeeze on the career stage that used to produce twenty years of experience.
Wisepreneurs Project podcast, Episode 38, on why overcharging produces immediate signals and underpricing produces almost none, and why clients rarely tell a supplier they are too cheap.
Auyon Siddiq and Niuniu Zhang - "Human Capital, AI, and Labor Commoditization," UCLA Anderson School of Management, 19 June 2026 (SSRN 6968139). The study at the centre of this article, providing the first empirical measurement of AI-driven labour commoditisation using five years of Upwork data, including the scope condition that bounds the finding to hiring through profiles and prices.
Aki Ito - "AI Is the Great Equalizer" (Business Insider). Reports Upwork research finding that top freelancers in image-based services took 7 percent fewer jobs and 14 percent lower earnings after generative AI arrived, evidence that the erosion hits the strongest performers hardest.
Ozge Demirci, Jonas Hannane and Xinrong Zhu - "Who Is AI Replacing? The Impact of Generative AI on Online Freelancing Platforms," Management Science (2025). Found a 21 percent fall in job postings for automation-prone freelance work after ChatGPT.
Every Tuesday I send The Wisepreneur, a short letter for experienced professionals building independent practice: one usable idea a week. Sign up at wisepreneurs.com.au/newsletter
When you want a second pair of eyes on your own practice, the Positioning and Practice Audit is how I work with people: wisepreneurs.com.au/marketing-partnership