'Welcome back to Terms of Service. I'm CNN tech reporter Clare Duffy. I don't think I need to tell anyone that life these days is expensive. Some people are now forgoing basic necessities like food, utilities, and gas because of a lack of affordability. If you've bought a carton of milk sometime in the past year, for example, you have seen this in action. But now imagine that the cost you're charged for a specific carton milk at a specific store is different from the cost your neighbor pays for the exact same product, maybe because you have different budgets or shopping habits. This is not a hypothetical, it's actually happening, thanks to a practice known as algorithmic pricing. It's also sometimes referred to as surveillance pricing or personalized pricing, and it's affecting online shoppers. It's a pricing strategy where AI-powered algorithms churn through consumer data, like where you live, what you make, and what you purchase. And also trends in consumer demand for certain products. Corporations then use that information to predict what you'd be willing to pay and adjust prices accordingly. To walk me through how this works and how it affects consumers, I'm talking to Grace Gedye, a policy analyst at Consumer Reports. She has some tips on how to look out for automated pricing schemes and what we can all do about it. My conversation with Grace, after this short break.

Hi Grace, thank you for being here.

Thank you so much for having me.

So how did you first become familiar with this issue of algorithmic pricing?

'Well, I shop around online like everyone else, and I've had the experience of seeing one price in one browser and being like, is this the best price? Opening a different browser, having a friend see a different price, and kind of wondering what's going on there. And through my work at Consumer Reports, we've been thinking a lot about how to protect consumers now that everyone's shopping online. And a big shift is everyone seeing the same price tag in the store to now everyone seeing prices on their private screens, and also companies having a ton of information about each of us individually. Everything from your search history to the type of device you're using, to your real-time location, battery life, but also demographic information, inferences about your income, all these things that can paint a really detailed picture of who you are, what you want, and how badly you want it.

'So can you explain what AI-driven algorithmic pricing is in layman's terms for us?

'Algorithmic pricing is kind of a catch-all phrase for a handful of different pricing strategies. The one I've been most focused on is personalized pricing, which, as you said, the price varies person-to-person based on something that company knows about you as an individual. So you and your neighbor and your sister might all see different prices for the exact same fancy hairdryer at the same store at the same time. There's also dynamic pricing where the price is changing over time based on, say, predicted demand and a couple other strategies.

Part of the reason there has been a lot of talk recently about technology enabling retailers to run price tests or do personalized pricing is because Consumer Reports published an investigation back in December into Instacart's pricing strategy that got a lot attention. Grace walked me through the methodology and how they compiled that report.

This was a big undertaking. We worked with two partner organizations, Groundwork Collaborative and More Perfect Union. And we also worked with over 400 consumer volunteers across the country to do these live price checks. And we were looking at Instacart, which is a grocery delivery platform. You can log onto the app and have someone pick out groceries from a major grocer on your behalf and then deliver them to you. And we were wondering, are people seeing different prices for the exact same item at the same store at the same time? And if so, what explains that? And at what cost? And so the investigative reporters put together these live Zooms where researchers guided consumers in logging into the app, picking out the exact same physical store, like I'm based in D.C., a Safeway in Washington D.C. was in the test. And then putting the exact same list of goods into their basket, taking a whole bunch of screenshots to record the prices. And then the researchers did a bunch of verification of those screenshots, a lot of data analysis, and found out that people were indeed seeing different prices for the exact same item at the same time.

And why did you choose to focus on Instacart?

Instacart, it seems, has invested pretty heavily in pricing technology and providing that technology to its retail partners. In the case of this investigation, they had purchased this company, Eversight AI. And so it seemed like they had a sophisticated operation that would perhaps enable this sort of variable pricing.

So what, ultimately, did you find here?

So, we found that for about three quarters of grocery items in our big test, there were different prices for the same goods at the same time for different people. And the highest spread we saw was an item where the highest price point was 23% higher than the lowest price point. We also saw some items with five different prices at the same time.

Wow.

So, like, a dozen eggs from the brand Lucerne in a DC Safeway had five different prizes. And this all adds up to, based on what Instacart says, a family of four spends on groceries in a year, a potential cost swing of $1,200 per year.

So you could be paying $1,200 more for the same products than your neighbor down the street is?

Yeah, over the course of a year.

Wow. So how does this work? What kinds of data are they using to determine different prices for different consumers?

'So, Instacart specifically said that they were doing randomized price testing, but there are lots of other examples of companies doing other forms of personalized pricing where investigative reporters have designed some really clever investigations to try and figure out what sort of information they're using. So, one example, investigative reporters in Minnesota were testing the Target app and found that when a customer was on the far side of a Target parking lot, so they were looking at a Dyson vacuum. They would see one price, they'd walk into the store with their phone, again, picking up real-time location data, and see the price of that vacuum jump significantly. Another example, Orbitz the travel website. The Wall Street Journal found that they were showing consumers different hotels at different prices based on whether they were searching from a Mac or a PC. People using Macs were being shown higher prices. A third example, Princeton test prep was found to be charging different prices for virtual tutoring based on the customer's zip code. So there are a whole bunch of types of examples, and as probably people listening to this podcast may well know now, companies have just a lot of data on each of us. So some of it is based on the device you're using, some of its kind of guesses based on your search history, your shopping patterns, and a lot can be inferred about you based on that information.

We should say that these companies, including Target and The Princeton Review, have previously said their prices are set to reflect different competitive factors in different markets. In response to the 2012 Wall Street Journal story on Orbitz that Grace mentioned, the company said at the time its price tests were "experimental." That's also how Instacart framed the pricing strategy highlighted in Grace's report.

Throughout the report, Instacart's pricing strategy is referred to as "experimental." That's sort of how the company has talked about this as well. But obviously, consumers didn't know they were participating in an experiment. And they paid for the findings of that experiment out of pocket, basically, with small price differences that, as you said, could add up to a big chunk of a family's budget over a year. How should we think about that? Just the fact that these companies are "experimenting," often without our knowledge?

In the Instacart example, they were actually bragging on a portion of the website that consumers didn't know this was happening. And I think part of the calculus here is when these examples have been revealed to the public in the past, they're met with outrage. People really do not like it. And so companies have an incentive to not make it clear what's happening. There have been lots of good reports, and now a law was recently passed in New York that requires a bit of disclosure. So I think more of this is coming to the fore. But even in the wake of the Instacart investigation, for example, we saw more than 12 members of Congress write letters either to Instacard or to the Federal Trade Commission, which is a consumer protection government agency. Reuters reported that the FTC was investigating Instacar for their pricing tactics and Instacart walked it back and said they would no longer be doing it. And we've seen other kind of instances of something coming to light and a company needing to walk it back.

'Yeah, I wanted to talk a little bit about Instacart's response here. They initially responded, as you said, by saying that each retailer's pricing policy was displayed on the platform so that customers can see if there are differences between online and in-store pricing, and said this was a limited subset of stores that were doing this price experimentation, they said, in order to keep prices lower for essential items for customers. But then a few weeks later, in the wake of this backlash, it said that it would end these price tests, meaning that going forward, all customers will pay the same thing for the same goods. What did you make of that response?

'I thought it was interesting and it does kind of suggest how strong public sentiment is on this issue. They did say they'd still be allowing partner grocers to test different promotions and discounts, which I find interesting because I do think the way retailers offer discounts has evolved a lot over the past 20 years and I do think discounts are now more personalized, more targeted and more complicated and less clear-cut when you're getting a good deal. And so, that's something that I'm kind of increasingly keeping my eye on is when is a discount real? When is it kind of approximating personalized pricing?

'I reached out to Instacart for comment on this episode. As Grace and I talked about, an Instacart spokesperson told me that its price tests were more random and not "based on personal demographic or user-level behavioral characteristics." But the spokesperson said Instacarte ended its tests after realizing that they upset consumers and said, "at a time when families are working hard to stretch their grocery budgets, customers should never have to question the prices they see on Instacart." Still, big picture, companies do have a lot of data on us, while consumers are given very little clarity into how prices are determined. Meaning, it can be hard to tell if they're equal and fair. So, how can personalized or algorithmic pricing affect your finances? And is there anything you can do about it? Grace has some tips. We'll be right back.

Do we have a sense of scale in terms of how many companies are doing this kind of algorithmically driven pricing?

'As kind of evidenced by like the scale of this Instacart investigation and what it took to pull it off, a tricky thing about it is it takes a lot of methodological rigor to really establish definitively that something is happening, and so that's resource-intensive. And so, most of the examples we have are from investigative journalism. That said, the Federal Trade Commission, this consumer protection agency, launched this study, and you know, government is in this unique position where it can compel companies to say what they are doing. And so, it sent a bunch of questions to a whole bunch of kind of middlemen-type companies that would offer price targeting products. So MasterCard, Accenture, JPMorgan Chase were among the companies that received these questions. And there were questions about, you know, how are they offering these price targeting products to clients, retailers? They learned that these practices were happening in a really wide array of retail segments from home goods to apparel to home renovation-type stores. It's not clear that that research is ongoing or if the final results will be revealed to the public.

You touched on this a little bit, but how does algorithmic pricing differ from other corporate pricing tactics that we've seen? Like, obviously retailers have argued that stores have done price testing for years, but how is this pricing different from what we've see before?

'There are many ways that retailers could test prices in stores 20, 30, 40 years ago. They could say, okay, in this store, we're going to vary these prices and see how much that changes the volume people buy. We're going mail out coupons and see that stimulates demand. The difference now is they know so much about us, way more than they did previously. You know, we did another investigation, investigation of Kroger's data practices. And under some consumer data protection laws, consumers have the right to request data that a company's collected about them. One of the consumers we worked with got back a 62-page profile with everything from inferences about his education level, his income.

Wow.

'The likelihood that he had a pet, his likelihood to travel. So this really fine-grained data and predictions that the store can then use or also sell to other stores.

'And is this personalized pricing happening mainly via online shopping platforms or are in-store shoppers potentially subject to this as well?

'It's certainly easier for a company to pull off online. In-store, typically we're all still seeing price tags, although certainly some retailers are starting to implement electronic price tags. One way that a similar phenomenon starts to happen in stores is these personalized discounts I referenced. You're being incentivized to use an app, get personalized discounts delivered to you, and functionally, that can mean that people are paying different prices for the same product. But certainly, the kind of variation on list prices I think we're more likely to see online.

Is there any potential upside for consumers here, like situations where some people are paying lower prices?

It's definitely possible that some people, in some cases, might end up paying a lower price, right? You know, if there's some backpack that you, Clare, really like, you'd be willing to pay $90 for it, but the retailers set a price at $50, you're $40 better off. And perhaps they know that I'd only be willing willing to pay $10 less than whatever I just said. So maybe I buy it and I'm $10 better off, but I think we can bet on companies is not rolling out this practice unless it results in them increasing their profits.

'Do consumers have protections against AI-driven algorithmic pricing?

Well, not as many as they should. It's not really clearly prohibited. So there are a couple different laws that perhaps intersect with it, right? There are consumer data privacy laws and depending on the type of information about your company is using, if it's particularly sensitive there might be some restrictions there, but by and large privacy laws don't prohibit this sort of thing. Some states have price gouging laws, but those are mostly linked to declared emergencies, like hurricanes or tornadoes, right? You don't want water being quintupled in price during a natural disaster. And then there are these kind of bigger, broader consumer protection laws that broadly prohibit unfair or deceptive practices. And perhaps there's an argument being made that this is unfair or deceptive, but I don't think we've seen a court case clearly establish that yet. So for now, it's not clear that it's illegal, and that's part of the reason that we at Consumer Reports and many other advocates think states and the federal government should pass some new laws to prohibit it.

So your investigation came out as we have seen some federal and state lawmakers attempting to curb strategies like personalized pricing. States including New York, Colorado, California, Georgia, Illinois, and Pennsylvania have all introduced or advanced legislation to limit individualized or algorithmic driven pricing. And at the federal level, Texas Congressman Greg Casar's 'Stop AI Price Gouging and Wage Fixing Act' would ban the use of personal data for individualized prices. What is the status of these policy updates, and do you actually think they could make a difference?

This is what I do day in and day out. I primarily, actually, work on state legislation because it tends to move a bit faster than Congress. And the list has even grown. Minnesota, Hawaii, there are all these states that are considering bills that would either require the disclosure of it or prohibit it. There's the New York law that I already mentioned that passed either last year or two years ago requiring disclosure. And that has even started bringing this issue to light a bit more. I don't think the sort of generic disclosure that the New York law requires is actually that helpful to the consumers. It's not clear what they should really do with the information. So I think prohibition is probably the right policy approach. And so many states are considering prohibitions, either for all goods and services or just one sector like grocery prices. And I think if one of those bills were to pass, that would really start to chip away this practice, build political momentum that policymakers can do something about it. And so I'm relatively optimistic. These political battles are always, always tricky and the lobbyists for the retailers and the tech companies are quite good at their jobs, but I think there's a lot of political appetite to do something about this problem.

The other thing that was so striking about the timing of your report is it comes at this time when food prices overall are outpacing inflation and Americans are reporting that the price of groceries is a major cost concern. What advice do you have for people shopping for basic necessities who don't want to end up paying more than their neighbor?

I'll share a couple pointers, but I do think the most important takeaway is it's not really reasonable for any individual consumer to try and beat the system here. There's a major information asymmetry. The company has a lot of information about you and you have no information about how they're doing pricing. So to a certain extent, the house always wins and I don't really think it's reasonable for individual consumers to be trying to kind of game this out on their own. That said, a couple things to think about. One is shopping in person when you can. And when you check out, if the store has a big loyalty program individual consumers on a treasure hunt to figure out don't think it's really reasonable, right? Like we should have baseline legal protections that people can expect that the price they're seeing is just the price and they haven't been profiled.

Well, Grace, thank you so much. This is so important and I think just really helpful for people to understand this, even if there's only so much they can do about it as individuals. So thank you.

Thank you, Clare.

'As Grace said, it's not practical to run your own investigation before you buy something online. But if you can, it may be worth shopping around and taking the extra step to compare prices. And consider shopping in person rather than online when possible. As we often advocate for on this show, it can also help to take basic steps to protect your private information online, like declining cookies and avoiding sharing location data with apps and websites whenever possible. There are also efforts at the state and federal level to pass legislation that would protect consumers from this kind of personalized pricing. So keep an eye out for that. Before we go, I want to update you on a bit of tech news in case you missed it. Back in November, we did an episode with AI expert Henry Eider about the explosion of AI slop content across the internet. It followed the launch of OpenAI's Sora app for creating and scrolling through AI-generated videos. Well, recently, OpenAI announced it is shutting down Sora and moving away from AI video generation to focus on other priorities. This isn't totally surprising. Fewer people were using Sora after the initial hype died down, and AI-generated video also requires a ton of expensive computing resources that OpenAI wants to use on products that could be more profitable. The app also suffered from criticisms about the use of intellectual property, and general backlash to AI-generated creative works. Now, plenty of other apps still offer AI video tools, but this is a pretty significant signal that one of the top players in this space has decided AI-generated video is not worth investing in. That's it for this week's episode of Terms of Service. I'm Clare Duffy, talk to you next week.