The advertising world’s next hot property: AI chatbots

A few weeks ago, I found myself planning a year-end trip using ChatGPT. I asked for destination recommendations, compared hotel options, and even got itinerary suggestions. The experience felt like having a travel agent at my fingertips. The chatbot provided relevant links to websites I could visit to complete my travel plans. As of now, however, the AI can recommend, but it cannot transact.

That gap is about to close. And when it does, the recommendations themselves may not be quite so neutral.

The platforms powering these conversations are now racing to introduce advertising into AI chatbots. Meta AI will begin using chat conversations to personalise ads starting December 16, with users potentially seeing relevant ads on Instagram and Facebook after asking the chatbot for recommendations. Elon Musk announced plans in August to introduce advertising into Grok's answers, allowing marketers to pay to appear in suggestions from X's AI chatbot. Google is testing ads in third-party AI chatbot conversations through its AdSense network, having begun trials with conversational AI search startups iAsk and Liner.

Even Sam Altman, OpenAI's CEO, recently acknowledged that though there are no current plans for ads, he didn't entirely rule out the possibility.

The conversational format is already proving its commercial value. "When we speak to our customers, we're already seeing click-through rates decline, even for top-ranking results. There's almost a 20-50% drop in traffic over a six-month to one-year period for certain keywords," explains Nishant Arora, SVP of Global Marketing at Netcore Cloud.

At the same time, traffic from AI Overview (AEO) and chat-based channels has “exploded”, according to him. “While it may represent a smaller share of total traffic today, the quality of that traffic is significantly higher. Whether in B2B or B2C contexts, visitors coming from chat interfaces show much stronger intent — conversion rates have increased by 2-3x compared to traditional CTR-based performance."

But there's a fundamental problem with this promising picture: most people aren't willing to pay for the privilege of talking to a chatbot. According to a Bloomberg Intelligence survey of 1,000 respondents conducted in May, only about 25% of users who use generative AI have paid subscriptions and are willing to pay up to $20 monthly for features including search, copilots, and image and video generation. OpenAI has more than 800 million weekly active users, but fewer than 20 million subscribe to paid plans.

This becomes particularly challenging given the funds going into maintaining the LLMs. Meta's CEO Mark Zuckerberg said in September that his company intends to spend $600 billion on building out AI infrastructure over the next few years. According to The Information, OpenAI is expected to spend more than $115 billion through 2029 on AI infrastructure. Anthropic has introduced a $200-a-month ‘Max’ tier. 

To say the least, these models are under pressure to find new revenue streams. Running LLMs at scale requires expensive data centre infrastructure packed with tens or even hundreds of thousands of high-end GPU chips, far more costly than the commodity servers that have powered traditional search advertising for decades.

Advertising emerges as the natural solution to this revenue gap. Arora notes that integrating advertising could expand the addressable market, with roughly $500 billion in potential ad revenue at stake. The industry is likely heading toward a tiered or freemium model, similar to Amazon Prime or Spotify, where basic access remains free but ad-supported, while users who want an uninterrupted experience pay for premium versions.

How ads will actually work in conversations

The advertising formats that will take shape could look nothing like traditional banner ads or search listings. "As platforms plan to embed ads natively within AI-driven chats, the distinction between content and commerce blurs; brands will appear within responses, product recommendations, and real-time dialogues, rather than as banners or static feeds," says Rupinder Singh, Founder and CBO of Thinkroi. "We can expect dynamic ad formats like sponsored suggestions within chat responses, shoppable cards and embedded mini-storefronts, interactive, context-driven product demos. These formats are adaptive, context-aware, and engineered for zero disruption — blurring the lines between user inquiry and brand moment."

The emerging formats are already being tested across platforms. Perplexity has experimented with sponsored follow-up questions that appear alongside answers to user prompts. Similarly, Chai, the romance and friendship chatbot, serves pop-up ads, with reports suggesting that users spend an average of 72 minutes daily on the platform.

Amropali Shetty, Director and Head of Global Marketing at Yellow.ai, notes, "Conversational interfaces are emerging as high-intent, bottom-funnel real estate where an ad is only valuable if it shortens the user's path to getting something done," she explains. The format is being normalised right now with tech platforms bringing in 'Sponsored' labels and follow-up questions that extend the conversation.

With chatbots likely bringing in ads, traditional targeting will give way to contextual targeting. In this scenario, relevance becomes a simple question: does this sponsored step reduce the user's effort in this exact moment? Platforms are experimenting with using conversational context to route ad products into answers, which means intent is now the entire thread, not a single search term, as per Shetty.

Think of it this way: if you ask a chatbot for holiday recommendations, a sponsored hotel suggestion appears while you’re actively planning. The nudge feels less like an ad and more like part of the conversation. These interventions show up inline within the answer or as clearly labelled follow-up prompts. The conversion impact is stronger because the timing and context align perfectly with user intent.

What happens to brand storytelling in three-turn exchanges

This shift forces a complete rethinking of creative strategy. While every brand aspires to have a certain personality — whether as a challenger, a thought leader, or a fun, approachable brand — what CMOs envisioned in the boardroom rarely translated perfectly into customer experiences. Arora explains, "With chat interfaces, that changes. Now, we can define exactly how our brand behaves during an interaction. A key question marketers will face is whether to adapt their brand personality to match the customer's tone or stay consistent with their established identity."

New technology now lets brands host AI agents across apps, websites, and even WhatsApp; agents that can initiate conversations, not just respond.

"The creative challenge isn't about making ads 'conversational' but about making them genuinely useful within the flow of what someone's already doing," Shetty explains. "What does that look like practically? A brand-safe micro-assistant that opens with clarity, asks one smart clarifying question if needed, and provides a concrete next action. You're not crafting a 30-second narrative arc anymore; you're designing a three-turn exchange that either solves the user's problem or gets out of the way."

This could lead to competition between AI agents themselves. When someone asks a chatbot to book a hotel, will the booking happen through Expedia's agent or Marriott's own brand agent? The same dynamic will play out across channels like WhatsApp or Meta's LLM. Users will choose the most useful brand, not just the most visible one, Arora shares.

The business models taking shape

The sustainable revenue models emerging from this shift blend multiple streams. “It could be auction-based search inventory flowing into chat (what's happening now), pay-per-resolution pricing for sponsored steps (where you pay when the user actually completes the task), and commerce or affiliate models where the conversation leads directly to purchase," says Shetty. 

She comments that the industry is moving toward outcome-based KPIs, not impression-based ones. 

Singh draws parallels to earlier platform shifts. Throughout his career, he has witnessed digital advertising evolve from Web 1.0 banner ads and basic click-based models to the transformation brought by Web 2.0 with platforms like YouTube and Facebook and Google's 2008 acquisition of DoubleClick that consolidated its dominance in digital advertising.

He says, “Today, we're seeing a similar pattern unfold with Perplexity's bold moves into the browser space through Comet. Much like Google's early acquisitions, Perplexity is not just launching a new product — it's actively shaping the future of how users interact with information and commerce online."

However, Perplexity recently pulled back on its advertising initiatives. Its head of publisher partnerships stated that the startup is not taking on new advertisers and that ads are not currently part of the plan for Perplexity’s browser, Comet. The move came shortly after its ad sales head departed, hinting at broader industry challenges in monetising AI search through ads.

Reports note that ad buyers have struggled to measure key metrics like click-through rates and return on ad spend because Perplexity lacks tools that match the standards of established ad platforms. Moreover, according to Attest, only 14% of consumers trust generative AI search results “a lot more” than organic search, and just 9% are highly confident in AI-generated results.

The question no one can afford to ignore

This brings us to the most fundamental challenge: if a chatbot recommends a product, can users ever know where algorithmic objectivity ends and advertising begins?

While Meta users won't be able to opt out of having their chats used for personalisation, conversations involving religious or political views, sexual orientation, health, and racial or ethnic origin won't be used to customise ads or content. Chats conducted before December 16 won't be used for personalisation, and the policy won't apply to users in the UK, South Korea, and the European Union, at least initially.

The privacy implications run deep. Many people are opening up to chat-based AI platforms, forming emotional connections through AI friends or companions as they share personal information. These systems hold the power to disguise ads as helpful content. The stakes around consent and disclosure are high.

"Users need to know why they're seeing something, and they need control; the ability to collapse, mute, or give feedback," Shetty emphasises. "The sponsored content must be clearly labelled, and critically, the underlying answer logic needs to stay independent from the commercial layer."

Pew Research shows users want more control and transparency around AI, and regulators are paying close attention. For example, India's Ministry of Electronics and Information Technology has proposed rules that require AI-generated content to carry visible labels covering at least 10% of the visual display area or 10% of the audio duration, with metadata traceability built into every piece of synthetic content.

On top of that, people turned to tools like ChatGPT because search engines like Google felt overloaded with ads on the first page. 

“If AI chat interfaces also blur the line between objective advice and paid content, users will lose that trust again," he says. "Another key factor is explainability. Most LLMs today can make mistakes, and unless the user challenges them, those errors go unchecked. LLMs must be able to explain the basis of their answers — showing their sources, how recent they are, and even their level of confidence."

Sam Altman articulated this tension directly, stating that ads on Google search are dependent on Google doing badly because if it were giving the best answer, there'd be no reason to buy an ad above it. He emphasised that if ChatGPT were accepting payment to put a worse hotel above a better hotel, that would be “catastrophic” for the relationship with users.

His proposed model shows one possible path forward: show the best recommendation first, then if a user books or buys with one click, take a small commission that doesn't affect results. 

The platforms experimenting with ads in conversational AI need revenue to sustain expensive infrastructure. They need scale to compete. But they also need trust, and once lost in this context, it may prove impossible to regain. The $500 billion opportunity Arora mentions is real, but only if the industry can solve the trust equation.