Every competitive intelligence programme runs on the same instinct: find out what the other side charges. Most teams treat that as a sourcing problem. It is also a legal one, and the line the law draws is not where most researchers assume it is.
Answer first
Asking your own customers and prospects what a competitor charges is lawful. Asking the competitor, or joining a scheme in which competitors feed each other price information, is where Section 1 of the Sherman Act starts to apply. The statute is one sentence: "Every contract, combination in the form of trust or otherwise, or conspiracy, in restraint of trade or commerce among the several States, or with foreign nations, is declared to be illegal."
The word doing the work there is conspiracy, and the surprise for research teams is how little it takes to establish one. You do not need a price agreement. You do not need a meeting, a memo, or an intent to fix anything. In the leading Supreme Court case on the subject, the entire violation consisted of competitors phoning each other to ask what they had most recently quoted.
That is not a hypothetical. It is a workflow, and it looks uncomfortably like the one in a lot of competitive intelligence plans.
The case where the research method was the violation
In United States v. Container Corp. of America (1969), corrugated container manufacturers accounting for "about 90% of the shipment of corrugated containers from plants in the Southeastern United States" had an informal habit: when one needed to know a rival's current price to a specific customer, it called and asked, and generally supplied the same courtesy in return.
The Court was explicit that no cartel had been proved. "There was here an exchange of price information but no agreement to adhere to a price schedule." What existed was only this: "Here all that was present was a request by each defendant of its competitor for information as to the most recent price charged or quoted, whenever it needed such information and whenever it was not available from another source."
The Court held that this alone was enough. The reciprocal expectation was "sufficient to establish the combination or conspiracy, the initial ingredient of a violation of" Section 1. It described the arrangement as "though somewhat casual" and still condemned it, because "The exchange of price data tends toward price uniformity," and because "Stabilizing prices as well as raising them is within the ban of" the Act.
Justice Marshall, dissenting, described what the participants actually got out of it in words that would fit neatly into a competitive intelligence brief: "In all cases, the information obtained was sufficient to inform the defendants of the price they would have to beat in order to obtain a particular sale."
Read that as a research requirement and it is exactly what a sales team asks for. Read it as a finding of fact in a Sherman Act case and it is the harm.
The majority's closing line is the one worth pinning above a discussion guide: "Price is too critical, too sensitive a control to allow it to be used even in an informal manner to restrain competition."
The distinction that actually decides it
Container Corp also contains the sentence that separates lawful market research from an unlawful exchange, and it is a distinction about the shape of the data, not about anyone's motives:
"There was here an exchange of information concerning specific sales to identified customers, not a statistical report on the average cost to all members, without identifying the parties to specific transactions."
Specific and identified is the problem. Aggregated and anonymous is the defence. Hold on to that, because it is the hinge of the whole area, and it has an awkward consequence we take up in the companion piece on benchmarking surveys and antitrust.
Nine years later, in United States v. United States Gypsum Co. (1978), the Court confirmed that information exchange is not automatically illegal: "The exchange of price data and other information among competitors does not invariably have anticompetitive effects; indeed such practices can in certain circumstances increase economic efficiency and render markets more, rather than less, competitive."
It then gave the test: "A number of factors including most prominently the structure of the industry involved and the nature of the information exchanged are generally considered in divining the procompetitive or anticompetitive effects of this type of interseller communication."
Two factors, then. Structure of the industry and nature of the information. Neither is about whether you meant well.
Gypsum is also a caution about defences that feel obviously reasonable. The practice at issue was "the practice of telephoning a competing manufacturer to determine the price being currently offered on gypsum board to a specific customer," and the sellers argued they were verifying prices in good faith to comply with the Robinson-Patman Act's meeting-competition defence. The Court of Appeals had treated that purpose as a controlling circumstance. The Supreme Court did not accept it as a blanket shield.
Where research teams actually get close to the line
None of this makes competitive research dangerous. It makes four specific patterns dangerous, and they are all avoidable.
Interviewing employees of a direct competitor. A prospect who happens to work at a rival is not a normal respondent. If the conversation moves to their pricing, terms, or forward plans, you are receiving competitively sensitive information from a competitor, which is the fact pattern the cases are about. Screen for employer and route those people out of pricing modules.
Trade association or peer-group data collection. A room of competitors comparing numbers is the classic setting. It can be done lawfully, but not casually, and not by a researcher improvising.
A shared intermediary. If the same vendor, consultant, or software product collects nonpublic figures from several competitors and hands back a number each uses to set price, the fact that nobody spoke directly to anybody is not the protection people assume. More on that below.
Asking a customer to hand over a competitor's document. A buyer's copy of a rival's quote or contract is often covered by a confidentiality clause. Receiving it creates a problem for them and a document for you.
Notice what is not on that list. Asking your own customers why they chose someone else, what they compared, what they think a fair price is, and what would make them switch is ordinary, lawful, first-party research. That is the overwhelming majority of what a competitive programme needs, and it is better evidence anyway. Our guide to competitive intelligence interviews and the broader competitive research guide both work entirely inside that boundary.
The 2025 development research teams missed
In January 2025 the Department of Justice and the Federal Trade Commission jointly issued Antitrust Guidelines for Business Activities Affecting Workers. Its first page states plainly: "This document replaces the Antitrust Guidance for Human Resource Professionals (2016)."
Two of its statements matter far beyond employment.
First, on intent: an exchange "may be unlawful when the information exchange has, or is likely to have, an anticompetitive effect, whether or not that effect was intended." Good faith is not the test. Effect is.
Second, on intermediaries: "Exchanging such information with competitors may be illegal even if companies use a third party or intermediary" - the guidelines add, "including a third party using an algorithm" - "to share such information." And more pointedly: "Information exchanges facilitated by or through a third party (including through an algorithm or other software) that are used to generate wage or other benefit recommendations can be unlawful even if the exchange does not require businesses to strictly adhere to those recommendations."
Non-binding recommendations from a shared tool, in other words, are not a safe design. Which brings us to the live case.
In United States v. RealPage, Inc., Civil Action No. 1:24-cv-00710 in the Middle District of North Carolina, the government's theory is that landlords sharing nonpublic data through a common pricing product aligned their prices. The litigation is still running: a proposed Final Judgment for one landlord defendant, Willow Bridge Property Company, was filed on 6 July 2026, on the allegation that its "agreements with RealPage, Inc. and other landlords to share information and align pricing violate Section 1 of the Sherman Act."
The remedy in that case is instructive about what regulators think safe data looks like, and it is a long way from real time. We take the numbers apart in the benchmarking piece.
Design the study so the question cannot go wrong
The practical fix is not a warning in a briefing document. It is study design, because the risk enters through improvisation: a moderator hears something interesting about a competitor's pricing and, doing exactly what good moderators are trained to do, follows the thread.
Three design decisions remove most of the exposure.
Fix the boundary in the instrument, not in the moderator's judgement. Decide in advance which questions are about the respondent's own experience, preferences and decisions, and never about a third party's confidential terms. Then make the instrument enforce it.
Use structured questions wherever the answer needs to be countable rather than explored. Koji supports six types - open_ended, scale, single_choice, multiple_choice, ranking and yes_no - and for competitive work the closed types are a compliance feature as much as an analysis one. "Which of these did you evaluate?" as a multiple_choice, "how did you rank them on value?" as a ranking, "would you consider switching?" as a yes_no: each is bounded. It cannot wander into a competitor's confidential rate card, because the response space does not contain one. Our notes on competitive intelligence surveys and the customer interview question bank show the pattern.
Screen on employer before the pricing module, not after. A single_choice screener that routes competitor employees away from sensitive sections costs nothing and removes the worst fact pattern entirely.
This is where an AI-moderated interview has a structural advantage over a human one, and it is not the advantage usually advertised. A human moderator is supposed to be curious, and curiosity is precisely the failure mode here. A Koji AI interviewer probes deeply within a brief you set and does not exceed it - no rapport-driven drift, no off-script follow-up because the respondent seemed willing. Every session is transcribed and analysed identically, so the scope of what was asked is auditable rather than remembered.
Compare that with the alternatives. Traditional panels and legacy platforms such as Qualtrics, SurveyMonkey or Typeform give you a static instrument with no probing at all, so depth costs you a human moderator and the drift that comes with them. Interview services like UserTesting and dscout put a person in the room. Repository tools like Dovetail analyse conversations after the fact, once whatever was said has already been said and stored. Koji is the only layer that gives you deep, probing conversation and a fixed, reviewable scope, because the moderator is an instrument you configure rather than a person you brief and hope. Add automatic thematic analysis and one-click reports and the competitive study that used to need an agency runs in a day.
A necessary caveat: this is background, not legal advice. Antitrust turns on market structure and specific facts, and the analysis differs outside the United States. If your programme involves competitors, trade associations, or a shared data intermediary, that is a conversation for counsel before fieldwork, not after.
Ask the people who actually compared you
Your customers evaluated your competitors, priced them, sat through their demos and chose. That knowledge is first-party, lawful and better evidence than anything you could extract from a rival. Most teams never systematically collect it because moderated interviews are slow and expensive.
Koji removes that constraint. Design the study, set the boundaries once, and run AI-moderated voice interviews with as many customers and lost prospects as you need - no scheduling, no moderator bias, no research expertise required. Themes and quotes are synthesised automatically, so you go from question to insight in hours rather than weeks. See win-loss interview questions for a starting instrument, or Porter's Five Forces to frame the analysis.
Start a competitive study with Koji and get the intelligence from the only source that is unambiguously yours to ask.
Frequently asked questions
Is competitive intelligence legal?
Yes, in the overwhelming majority of forms. Gathering information about competitors from customers, prospects, public sources, and your own sales team is ordinary lawful business research. The legal risk under Section 1 of the Sherman Act arises from exchanges with competitors - directly, through a trade association, or through a shared intermediary - particularly when the information is current, nonpublic and price related.
Can I ask a customer what my competitor charges them?
Generally yes. The customer is not your competitor, and their own purchase price is their information to discuss. Two cautions: the price may be covered by a confidentiality clause in their contract with the vendor, and you should not ask them to send you the rival's quote or agreement. Asking what they paid and how they judged the value is different from soliciting a competitor's document.
Does it matter that we never intended to affect prices?
Under the 2025 DOJ and FTC guidelines, an exchange may be unlawful "whether or not that effect was intended." Intent can matter to other questions, including criminal exposure, but a sincere research motive is not by itself a defence to an information-exchange claim.
What if a third-party vendor collects the data, so competitors never talk?
That structure helps, but it is not automatically sufficient. The 2025 guidelines state that an exchange may be illegal "even if companies use a third party or intermediary" to share the information, including one using an algorithm, and even where any resulting recommendation is non-binding. The RealPage litigation is built on exactly that shape of arrangement.
Are win-loss interviews affected by any of this?
Almost never. Win-loss research asks your own buyers about their own decision, which is first-party research with no competitor on the other side of the table. Keep it that way by screening out respondents who work for a direct competitor and by not requesting rival documents. See win-loss analysis for the method.
How does an AI interviewer reduce this risk compared with a human moderator?
A human moderator improvises, which is their value and, here, their hazard: an unprompted disclosure about a competitor's pricing is the moment a skilled interviewer instinctively pursues. A Koji AI interviewer probes only within the brief you configure, applies the same scope to every session, and produces a complete transcript of what was asked. The boundary is enforced by the instrument rather than recalled by a person.
