The FTC wants to regulate AI answers - why 'steered' results could be illegal

Most of what shapes AI search comes from the platforms. This came from a regulator. On 7 July 2026 the FTC proposed that when an AI system is quietly steered toward a goal the user did not expect, that may be illegal deception. It is the first serious US signal that AI answers will be held to account - and the direction matters more than the specifics.

Regulation of AI search was always coming. What is notable is the angle the FTC chose: not copyright, not privacy, but deception - the idea that an AI answer quietly bent toward an undisclosed goal misleads the user who trusted it.

What was actually proposed

On 7 July 2026 the FTC published a proposed policy statement on the suppression of accuracy in AI systems. The core idea: if an AI company steers or distorts its system's outputs toward an objective the user was not told about - a commercial goal, a hidden preference, anything other than the honest answer the user expected - that can amount to deception under Section 5 of the FTC Act. The public comment period is open until 31 July 2026.

It is a proposal, not a final rule, and it will draw heavy comment. But proposals like this set the frame regulators and courts reason from, so the direction is the signal.

"An AI answer quietly bent toward an undisclosed goal misleads the user who trusted it."

Curious how AI engines describe your brand right now? Get a free visibility audit and see where you stand across ChatGPT, Gemini and Perplexity.

Why it matters beyond the legal detail

The AI answer sits in a position of unusual trust. A user reads one synthesised response and acts on it, often without checking the sources. If that response can be steered - toward a paying partner, a preferred product, a particular framing - without the user knowing, the harm is obvious. The FTC is proposing to name that harm.

For anyone building visibility in AI, the second-order effect is what counts. If undisclosed steering becomes a legal risk, engines face pressure toward transparency about how they select and rank sources. Today that process is a black box you optimise against by inference. A more accountable, more disclosed process is one you could actually understand and work with.

What brands should take from it

• Expect more accountability, not less. Between this and moves like Germany's media-law ruling, the trend is toward treating AI answers as things someone is responsible for. Plan as if the answer will be scrutinised.
• Clarity and verifiability gain value. A scrutinised engine prefers sources it can defend. Accurate, attributable, well-documented facts about you are exactly what survives that scrutiny.
• Watch the disclosure question. If regulators push engines to reveal how they select sources, that disclosure is the closest thing to a rulebook this space has had. Being early to read it is an edge.

The takeaway

The FTC proposal is early and contested, and it may change shape before anything is final. But it is a clear statement of an idea that is spreading: an AI answer is not a neutral act, and quietly steering it can be a harm the law recognises. For brands, that future rewards the same thing good AEO already rewards - being the clear, credible, verifiable source an accountable engine is safe to cite.

What 'steering' actually looks like

The word steering sounds abstract until you picture it. Imagine a user asks an assistant for the best project-management tool for a small team. The honest answer weighs the field on merit. A steered answer quietly moves a paying partner to the top, drops a stronger rival that never signed a deal, or frames the whole category so that one product looks like the obvious default. The user reads a confident paragraph and never sees the thumb on the scale.

That is the harm the FTC is circling. It is not the presence of a commercial relationship that is the problem - plenty of honest systems carry ads and say so. It is the gap between what the user was led to expect and what they actually got. A search page with a labelled 'Sponsored' row is disclosed. A synthesised answer that reads as neutral judgement while a hidden objective bends it is not.

For a brand, the useful mental model is simple. If your visibility in an engine depends on the engine quietly distorting an answer in your favour, that visibility is now sitting on legal quicksand. If your visibility depends on being the most accurate, best-documented, most citable source on the question, it is sitting on the exact ground regulators are trying to protect.

What this does not mean

It is worth being precise, because a proposal like this is easy to over-read. This is not a ban on AI answers, not a ban on commercial partnerships inside them, and not a rule that engines must recommend everyone equally. Ranking is not deception. An engine choosing to name three sources and not thirty is doing its job, not breaking the law.

Nor does it hand brands a lever to demand inclusion. Nothing here says an engine must cite you, and nothing here lets you argue that being left out is unlawful. The proposal is aimed at the engine's honesty with the user, not at the engine's obligations to any particular company hoping to appear.

And it is a US move. It carries weight because the FTC sets a tone others follow, but it does not by itself bind engines outside American jurisdiction. Read it as the leading edge of a direction, not as a settled rulebook you can plan a quarter around.

"If your visibility depends on the engine distorting an answer in your favour, that visibility is now sitting on legal quicksand."

How to position for it now

There is nothing here to action in a panic, but there is a posture worth adopting early. An engine under scrutiny becomes conservative. It leans on sources it can defend if a regulator ever asks why it said what it said. The work, then, is to be that defensible source before the pressure arrives.

• Make your claims checkable. Put the numbers, dates and specifics an engine can verify in plain text on your own pages, not buried in a PDF or a graphic. A fact an engine can confirm is a fact it is safe to repeat.
• Keep your facts consistent everywhere. When your pricing, positioning and product claims say the same thing across your site, your profiles and third-party mentions, an engine has less reason to hedge and more reason to name you plainly.
• Watch the disclosure question closely. If regulators eventually force engines to reveal how they select and rank sources, that disclosure is the nearest thing this field has ever had to a published rulebook. Being early to read it is a real edge.

None of this is new advice, which is the point. The FTC proposal does not change what good visibility work looks like - it raises the value of doing it well. A more accountable engine rewards the same clarity, accuracy and verifiability that already win citations today.