Perplexity says you cannot advertise to the machine

An early experiment in AI advertising has produced the first clear line in the sand. Time has been placing FAQ-formatted, brand-approved Agent Ads inside markdown versions of its webpages, versions intended for AI agents rather than ordinary human readers. Perplexity has now blocked those ads from influencing its search index and warned that publishers using what it calls deceptive advertising could suffer a reputational downgrade in the index. The disagreement is not a minor dispute over an ad format. It is an argument about the rules of AEO. [1] [2]

The first buyers included Ally Bank and the Project Management Institute. Time and its ad-technology partner Mobian presented the format as a way to give answer engines current, cited and verifiable facts about a brand. Perplexity sees the same design differently: a page where software receives information that people do not see is, in its view, a form of cloaking. [1]

<h2>What Time was trying to sell</h2>

Time had already been converting its webpages into stripped-down markdown versions, making the content easier for AI systems and agents to read. Mobian then inserted sponsored, FAQ-like material based on an advertiser's brief into those bot-readable pages. The material was labelled sponsored content, reviewed by the advertiser and designed to be contextual to Time's editorial franchises. The commercial proposition was straightforward: a trusted publisher's machine-readable environment could become a way for brands to influence the information an AI system retrieves. [2]

That proposal has a surface logic. Answer engines regularly need current information about pricing, products, skills, availability, regulations and brand claims. Structured data, product feeds, APIs, XML sitemaps and machine-readable documents have all long been designed partly for software. The fact that an information asset is useful to a machine does not make it deceptive.

<h2>Why Perplexity objects</h2>

The difficulty begins when the material is available in one version of a page but not another. Perplexity's Jesse Dwyer told Business Insider that serving one page to people and a different one to their software is cloaking, a practice search engines have historically penalised. Digiday reported that Perplexity has blocked markdown advertising on Time.com from influencing its agents and user-facing answers, and that its search and security teams are considering broader protections. [1] [3]

The platform's concern is not simply that a claim might be false. A paid claim can be factual and still create a problem if the system subsequently presents it as neutral, publisher-backed information. A sponsorship label may exist in markdown, but it may not survive into the model's final answer, citation or recommendation. The user could receive an apparently independent statement without knowing that the information originated in an advertiser's brief.

<h2>Disclosure alone may not settle the question</h2>

Mobian has argued that the ads are clearly labelled, verifiable and available for a model to use or ignore. That is a stronger position than fake reviews, covert brand mentions or invented customer conversations. Those tactics manufacture the appearance of independent support and should sit plainly outside legitimate AEO. The Time experiment is more difficult because the material is identified as advertising and can contain accurate source citations. [1]

But accurate information and legitimate influence are not necessarily the same thing. If an advertiser can buy a place inside the machine-readable layer of a publisher's site, the relevant question becomes whether the AI system can reliably preserve the commercial context when it evaluates, cites and communicates that information. Perplexity has answered no, at least for this format. Its decision matters because the platform has not merely ignored the ads. It has said the practice may affect how it evaluates the publisher itself. [3]

<h2>The three forms of AI representation</h2>

The dispute makes a useful distinction for the emerging AEO market. Earned representation occurs when an answer engine retrieves ordinary, credible sources and independently chooses to use them. Paid representation occurs when commercial information is introduced into an AI environment with clear disclosure and rules that allow the system and the user to understand its status. Manipulated representation occurs when commercial influence is disguised as independent evidence, including fake consumer conversations, hidden promotional material or a different factual environment for machines and people.

The first category is the foundation of durable AEO. The second will almost certainly grow through commerce partnerships, local lead formats and agentic advertising products. The third is where the category becomes corrosive. The challenge is that the boundary between paid and manipulated representation cannot be defined only by whether a human can see a page or whether a brand's claims are technically true. It depends on whether the source, sponsorship and evidentiary status remain intelligible throughout the retrieval-to-answer chain.

<h2>AEO needs an evidence-integrity standard</h2>

A practical standard would require four things. First, the factual core of information available to an AI should be available for human inspection as well. Second, any paid material should carry machine-readable disclosure that persists wherever the information is cited or summarised. Third, the claims should be independently verifiable through primary evidence, not merely asserted by the advertiser. Fourth, platforms should disclose how commercial material is treated in retrieval, ranking and recommendation. These principles would not prohibit machine-readable advertising. They would prevent it from borrowing the authority of independent evidence without accountability.

OpenAI and Google declined to tell Business Insider whether they would follow Perplexity's approach, while Anthropic did not respond. Time said it wanted a constructive discussion about the standards and safeguards that would allow advertising and AI to develop responsibly together. Those unanswered questions are the story. In SEO, search engines eventually created rules that separated useful optimisation from manipulation. AEO is now beginning the same process. [1]

<h2>AEO Updates Takeaway</h2>

Perplexity's decision does not settle the ethics of advertising to AI agents. It establishes that the platforms themselves will decide which forms of machine influence they consider legitimate, and may enforce those decisions before an industry standard emerges. Brands should not interpret this as a reason to stop making information machine-readable. They should interpret it as a reason to make claims, evidence and disclosures more durable. The asset worth building is not a hidden message for a crawler. It is a transparent, verifiable information environment that can withstand scrutiny by the model and the person relying on it.

<h3>References</h3>

[1] Business Insider, <a href='https://www.businessinsider.com/perplexity-thwarts-times-ads-for-ai-bots-experiment-2026-8' target='_blank' rel='noopener noreferrer'>The dream of serving ads to AI agents has hit a snag</a>, August 12, 2026.

[2] Digiday, <a href='https://digiday.com/media/time-has-started-serving-ads-to-ai-agents/' target='_blank' rel='noopener noreferrer'>Time has started serving ads to AI agents</a>, July 30, 2026.

[3] Digiday, <a href='https://digiday.com/media/perplexity-blocks-times-ads-served-to-ai-agents-calling-them-deceptive/' target='_blank' rel='noopener noreferrer'>Perplexity blocks Time's ads served to AI agents, calling them deceptive</a>, August 11, 2026.