Algorithmic pricing: Where is the legal line for a retailer?
.avif)
November 11, 2026
Dynamic pricing is permitted in France. Personalized pricing based on automated decisions is allowed, but customers must be notified of it. Algorithm-based price-fixing remains prohibited, even in the absence of contact between competitors.
The strongest safeguard for a retailer: being able to say, for each price, which rule generated it, what data it was based on, and who approved it.
The same algorithm can lower a price because a competitor has made a move, raise it for a specific customer, or align it with a rival’s price—all without anyone communicating with one another. Three practices, three legal regimes.
This article is intended for retailers’ pricing, sales, and legal departments. It outlines the distinctions between these three areas under French and European law, and then explains why the traceability of each price is the best form of protection. This is not legal advice.
Algorithmic Pricing: Three Practices, Three Models
Algorithmic pricing refers to the setting or recommendation of prices by a program, based on rules, data, and sometimes AI models. The law does not judge it as a whole; it examines what causes the price to vary and for whom.
Dynamic pricing varies depending on the market: demand, inventory, season, and competitors' retail prices. This is the principle behind dynamic pricing: at any given moment, all customers see the same price.
Personalized pricing varies depending on the individual: their history, browsing behavior, and profile. Finally, algorithmic coordination brings competitors' prices closer together through a shared tool or shared data.
| Criterion | Dynamic Pricing | Custom Price | Algorithmic Agreement |
|---|---|---|---|
| Factors That Affect the Price | The Market: Demand, Inventory, and Competitors' Retail Prices | The profile of an identified customer | Data or a tool shared among competitors |
| The same price for everyone at the same time | Yes | No | Yes, but in a coordinated manner |
| Senior Manager | Price Freedom, Rules for Price Display and Advertising of Discounts | Customer Information (Omnibus), GDPR | Antitrust Law (Article 101 TFEU, L. 420-1) |
| Status | Authorized | Subject to certain conditions | Prohibited |
The line between the first two categories is thinner than it seems. A price that varies by store is part of a dynamic pricing strategy. A discount calculated based on the profile of an identified customer is part of a personalized approach.
Dynamic pricing is allowed, as long as it remains legible
In France, prices are freely determined by market competition (Article L. 410-2 of the Commercial Code). The fact that an algorithm adjusts a price to market conditions does not alter this principle.
The Omnibus Directive explicitly states this in Recital 45: the required disclosure regarding personalized pricing does not apply to “dynamic” or “real-time” pricing that does not involve personalization.
"Free" does not mean "without rules." The listed price must be accurate, and a sale announcement must comply with the 30-day reference price, even when it is triggered by an algorithm.
The risk here is primarily operational: a price that changes faster than the label or product description. That is the whole point of well-coordinated omnichannel dynamic pricing.
Custom pricing is allowed, as long as you disclose it
Effective May 28, 2022, the Omnibus Directive—transposed into French law by Ordinance No. 2021-1734—requires that consumers be informed when a price is personalized for them based on automated decision-making.
The text does not prohibit the practice. It requires transparency with regard to the information that must be provided before a distance contract is concluded—that is, primarily online.
The GDPR adds to this rule. A price calculated based on a person’s profile constitutes profiling, which entails certain information obligations and, in the case of a fully automated decision, the safeguards set forth in Article 22.
This practice is by no means theoretical. In the United States, the FTC released the initial findings of its study on “surveillance pricing” in January 2025.
At least some customers—from grocery shoppers to clothing buyers—were using the intermediaries studied by the FTC, which rely on location, purchase history, or browsing behavior to tailor prices and discounts.
(FTC, Study 6(b) on Pricing Monitoring, Preliminary Findings, January 17, 2025)
New York has gone a step further. Since November 10, 2025, any price set by an algorithm based on personal data must include the statement “this price was set by an algorithm using your personal data.”
European law does not require such a direct approach, but the general direction is the same. An individualized price that the retailer cannot explain becomes a legal and reputational risk, as noted in our fact sheet on price discrimination.
Algorithmic collusion remains prohibited, even without a meeting
This is the only area where there is no room for interpretation. Article 101 of the TFEU and Article L. 420-1 of the Commercial Code prohibit price-fixing agreements, whether they take place during a meeting, via email, or through software.
The European Commission stated this in its 2023 guidelines on agreements between competitors. A company cannot evade its liability on the grounds that its prices were set by an algorithm that remains under its control (paragraph 379).
The same text addresses the situation in which competitors use a pricing tool provided by a third party that aggregates their sensitive information. According to the Commission, this could amount to horizontal collusion (paragraphs 401 and 402).
The RealPage case provides a concrete example of this. The U.S. Department of Justice accused this rental software company of using non-public data from competing landlords to recommend prices to them.
minimum age requirement for non-public data that RealPage may use to train its models, according to the agreement proposed by the U.S. Department of Justice, which is still pending before the court.
(U.S. Department of Justice, press release dated November 24, 2025)
In France, the Competition Authority lists algorithmic collusion among the risks associated with the automation of actions by AI agents in its Opinion 26-A-05 dated July 18, 2026.
Monitoring a competitor’s public prices remains legitimate: that is the very principle of price monitoring. What crosses the line is the exchange of non-public data, or a coordinated pricing strategy rather than one decided unilaterally.
What the AI Act Changes for Retail Pricing—and What It Doesn't Change
The European AI Regulation (Regulation 2024/1689) does not classify retail pricing as a high-risk system. The only pricing mentioned in Annex III concerns life and health insurance, alongside solvency assessment.
However, as of February 2, 2025, it prohibits systems that exploit a person’s vulnerability—whether due to age, disability, or social or economic circumstances—to alter that person’s behavior in a way that causes significant harm.
An algorithm that recommends prices by product and by store does not fall into these categories. A personalized price that exploits a customer’s vulnerability might come close—all the more reason to keep all personalization under human control.
The issue becomes more acute withagent-based pricing, where AI itself makes certain decisions. But the AI Act is not the legislation that governs retail pricing: that role is played by consumer protection and competition laws.
Traceability of each price, better protection for retailers
The three areas have one thing in common: When dealing with a customer, an audit, or a complaint, you’ll need to explain how a price was set—sometimes even months later.
For a retailer, this means being able to answer three questions about any price charged.
What rule?
The rule or objective applied: market alignment, target margin, psychological threshold, product line constraints.
What data
Input data: sales, inventory, competitors' retail prices, and no data that does not need to be included.
Who approved it?
The person or organization that accepted the award, along with the date and any exceptions.
Source: Booper, analysis framework, November 2026.
A price that answers these three questions stands up to scrutiny. It shows that it is determined by the market rather than by a specific profile, by public data rather than by an exchange between competitors, and by a human decision rather than by a black box.
That is why we viewthe explainability of pricing recommendations as a compliance asset, not just a convenience for teams.
Governance is just as important as the algorithm: who can deviate from the pricing policy, within what limits, and how is this documented?
In public debate, algorithms and fairness are often pitted against each other. For a retailer, the real dividing line lies elsewhere: between prices for which the decision-making process can be reconstructed, and those for which it cannot.
A high-performing but opaque ad server exposes the brand across all three regions at once. A simpler but traceable ad server protects the brand across all three.
Three Questions to Ask Your Pricing Solution
Without conducting a full legal audit, three questions are enough to assess a tool.
- What data is used to calculate the price, and does any of it describe an identified customer?
- Are the competitor prices used publicly available, or do they come from a shared database contributed to by other retailers?
- For a past price, is it possible to find the rule that was applied, the input data, and the person who approved it?
If you receive a vague response to any of these three questions, you should discuss the matter further with your editor and legal counsel. The same standard of proof applies to the accuracy of listed prices.
This article provides a general overview of the law in effect as of November 11, 2026. It does not constitute legal advice: for a specific case, consult your legal counsel.
Would you like prices that you can explain, one by one?
Our experts show you how each recommendation retains its rules, data, and validation, ready for an audit.
Let's discuss your pricing strategyFrequently Asked Questions About Algorithmic Pricing
Algorithmic pricing refers to the setting or recommendation of prices by a computer program, based on rules, data, and sometimes artificial intelligence models. It encompasses a wide range of practices.
The law distinguishes primarily between three cases: dynamic pricing, which follows the market and remains unregulated; personalized pricing, which is calculated for an individual and subject to disclosure requirements; and algorithmic collusion, which coordinates competitors’ prices and remains prohibited.
Yes. In France, prices are freely determined by market competition, in accordance with Article L. 410-2 of the Commercial Code. It is therefore permitted to adjust prices based on demand, inventory levels, or competitors’ publicly listed prices, including through the use of an algorithm.
The usual rules still apply: accurate price display, consistency between the price tag and the register, and adherence to the 30-day reference price for any discount promotion.
Yes, under certain conditions. Since May 28, 2022, the Omnibus Directive, which has been implemented in France, requires that consumers be informed when a price is personalized for them based on automated decision-making, particularly in online sales.
The GDPR also applies, since such pricing is based on profiling. Finally, personalization that exploits a customer’s vulnerability would expose the retailer to liability under fair trade laws and, in extreme cases, to the prohibitions of the AI Act.
Dynamic pricing varies depending on the market: demand, season, inventory, and competitors' prices. At any given moment, it is the same for all customers in a given channel or store.
Personalized pricing varies depending on the individual—based on their purchase history, browsing activity, or profile. It is this personalization that triggers the disclosure requirement under the Omnibus Directive, which expressly excludes dynamic pricing without personalization.
An algorithmic cartel is a form of price coordination among competitors that relies on an algorithm: a shared tool that aggregates their non-public data, or programmed rules designed to align their prices in concert. It is prohibited in the same way as a traditional cartel.
In its 2023 guidelines, the European Commission reiterates that a company cannot evade its responsibility on the grounds that its prices were set by an algorithm. Simply following a competitor’s publicly listed prices, however, remains legitimate.
The European AI Regulation (2024/1689) does not classify retail pricing as a high-risk system: the only pricing covered by Annex III relates to life and health insurance, specifically in connection with solvency assessments.
However, as of February 2025, it prohibits systems that exploit a person’s vulnerability to alter their behavior to their detriment. For retail pricing, consumer law and competition law remain the governing legal frameworks.
By being able to reconstruct the decision. For each price applied, it must be possible to identify which rule or objective led to it, what data informed it, and who approved it, along with the date and any exceptions.
This data shows that the price depends on the market and not on a profile, on public data and not on an exchange between competitors. That is why the explainability of recommendations and the governance of validations are becoming compliance assets.
An AI agent does not simply look at a listed price: it compares the actual price—which includes applicable promotions, shipping, availability, and unit price—for the products the customer requests.
Competitiveness is now measured rather than perceived. The accuracy and consistency of pricing data across channels have become the brand's new calling card.
An LLM can match differently worded descriptions and suggest groups of products from the same line. It can also be confidently wrong.
Booper's AI-driven chaining system combines AI, programmed controls that correct its errors, and human validation. No group is created without the user's consent.
An LLM integrated into pricing software processes sensitive data: prices, margins, and volumes. Where the model runs determines who has access to it.
Self-hosted AI keeps the data with the vendor or the client and keeps costs fixed. It does not eliminate the need for checks on each response or for human validation.