Is Perplexity's AI Agent Making Ad Agencies Obsolete?

A Signal from Social Media

Recently, a tweet from X sparked discussion in both marketing and tech circles. The claim was brief: "Perplexity's Agents are now more effective at running ads, for merchants, than advertising agencies."

The statement came with no data, case studies, or screenshots, but it points to an increasingly clear trend: AI Agents are moving from the "demo stage" into "actual business operations," and the first to be disrupted may be traditional ad agency services.

Why Advertising?

Ad buying is a classic "data-intensive + high-frequency decision-making + quantifiable feedback" scenario—exactly the three things AI Agents are best at handling.

The traditional ad agency workflow roughly includes: audience research, creative ideation, budget allocation, A/B testing, and performance review. A large portion of these steps relies on human judgment and experiential intuition, with long cycles and high costs. An Agent, by contrast, can continuously monitor ad performance, automatically adjust bids and audience targeting, rapidly generate and iterate creatives, and never needs rest.

Perplexity itself is known as an AI search engine that "can search the web and give answers with sources." Its Agent capabilities naturally include information gathering, comparative analysis, and task orchestration. Applied to advertising, this means it can more quickly find competitor strategies, industry benchmarks, and audience preferences, and translate conclusions directly into executable actions.

Perplexity and the Positioning of AI Agents

Perplexity AI is a company that started as a conversational search engine. Its core product retrieves real-time information from the web and generates answers with cited sources, distinguishing it from traditional large models that rely solely on training data. Around 2025, Perplexity introduced Agent capabilities (such as the Comet browser and task agents), enabling the system not only to answer questions but also to autonomously execute multi-step tasks, including information gathering, comparative analysis, task orchestration, and even calling external tools to complete operations. This combination of "retrieval + reasoning + execution" gives it a natural advantage in scenarios like ad buying that require continuous access to market intelligence and rapid response. It should be noted that Perplexity has not publicly launched a standalone ad-buying agent product, and the specific form referred to in the tweet remains unclear. Readers should view it as a capability demonstration or third-party use case rather than a mature commercial product.

The Real Differences Between Agents and Ad Agencies

It's important to be clear that friction between business owners and ad agencies has long existed: high agency fees, long communication chains, and slow responses to the needs of small and medium-sized merchants. The appeal of AI Agents lies in dramatically compressing this intermediary cost.

However, claiming Agents are "better" still lacks public quantitative evidence. The tweet itself provided no specific ad channels, budget scale, or conversion data. Ad agencies provide not just execution but also brand strategy, creative direction, and crisis judgment—all of which still require human oversight. A more accurate understanding is: for standardized, performance-oriented ad tasks, Agents have clear advantages in execution efficiency and marginal cost; but in brand building and complex decision-making, the role of humans and agencies has not yet been replaced.

Ad Agency Pricing Models and Intermediary Costs

Traditional ad agencies typically earn revenue from two sources: media agency fees (a percentage of ad spend, with industry practice around 10%–15%) and service fees (planning, creative, and optimization fees charged monthly or per project). In addition, agencies have multiple layers of roles internally—account executives, media buyers, optimization specialists—and communication and coordination costs ultimately get passed on to advertisers. For small and medium-sized merchants with limited budgets, this structure means a significant proportion of spending does not go directly toward ad exposure itself. AI Agents provide similar execution capabilities via subscription or usage-based pricing, theoretically able to significantly compress this intermediary cost. But note that agency fee percentages vary widely by market, channel, and contract type, and some platforms (such as performance advertising) already offer automated ad tools themselves—the incremental value of Agents needs to be evaluated in the context of specific channels.

What This Means for Merchants

For small and medium-sized merchants, the most direct value of this shift is lowering the barrier to professional ad buying. Daily tasks that previously had to be handed to agencies can now be handled by a subscription-based Agent tool, allowing business owners to put more budget into creatives and the product itself.

But stronger tools also mean judgment becomes more important. An Agent can execute, but it doesn't necessarily understand your brand boundaries and long-term goals. Business owners need to learn to define objectives, review outputs, and set budget red lines—these capabilities are becoming the new fundamentals of running a business.

The Fragmentation of the Ad Industry

If such tools continue to mature, the ad industry may stratify: basic execution and ad optimization will be largely taken over by AI, while agencies' value shifts upward toward strategic consulting, creative production, and integrated marketing. This is not an overnight replacement but a migration curve that has already begun.

Whether Perplexity has truly surpassed ad agencies in ad buying is still just a claim from a single tweet. But the question it raises deserves serious attention from everyone in marketing: when an AI Agent can research on its own, make decisions, and adjust budgets, what exactly are you paying an agency for?

Existing Practices of AI in Advertising

AI-assisted ad buying is not entirely new. Google Ads' Smart Bidding, Meta's Advantage+ automated campaigns, and various programmatic buying platforms have long used machine learning to optimize bids, audiences, and creative combinations. The addition of large language models and Agents mainly expands capabilities in "unstructured information processing" and "cross-platform task orchestration"—for example, automatically analyzing competitors' ad creatives, generating multiple copy variants, and adjusting strategies based on public sentiment. Therefore, a more accurate assessment is: the migration curve of automated ad buying began long ago, and Agents are an accelerator of this curve, not its starting point. Evaluating the actual effectiveness of Agents should focus on their performance in creative generation quality, cross-channel coordination, and exception handling, rather than simply comparing "human vs. machine."