Paid advertising just shifted. Three days ago, OpenAI announced that ChatGPT will begin testing ads within its free tier and newly launched $8-per-month Go subscription. This follows Perplexity’s move in November 2024 to introduce sponsored placements in its AI-powered search responses. The landscape is changing fast, but here’s the reality: nobody, including us, knows exactly how this will play out.
We’re observing developments closely, strategising with clients, and preparing to test when access opens. This isn’t a proven playbook. It’s early-stage thinking about how paid advertising might work in AI platforms. Here’s what we know so far, what we’re seeing, and how we’re preparing.
What is LLM advertising?
LLM advertising is the placement of paid ads within AI-generated responses on platforms like ChatGPT, Perplexity, Google Gemini, and others. Unlike traditional search ads that appear as separate sponsored results, these ads integrate directly into conversational AI outputs, appearing as sponsored follow-up questions, contextual recommendations, or labelled placements alongside answers.
The shift is significant. Google Search ads rely on keyword targeting and appear in distinct sponsored sections. LLM ads respond to conversational context and intent, appearing within the flow of AI-generated content.
Here’s how the two compare:
| Traditional Search Ads (Google) | LLM Advertising |
|---|---|
| Separate sponsored results section | Integrated within AI responses |
| Keyword-based targeting | Contextual, conversational targeting |
| Users ready to compare and buy | Users exploring and learning |
| Pay-per-click (CPC) standard | CPM and hybrid models emerging |
| Multiple ads per page (4-7+) | Low ad load (1-2 per session) |
Why now? Running large language models is expensive. ChatGPT reportedly costs around $700,000 per day in compute infrastructure. Subscriptions help, but don’t cover costs for free-tier users. Platforms need revenue. As DEPT Agency notes, every digital platform that reaches massive scale eventually faces the same gravitational pull: monetisation.
The scale is enormous. ChatGPT has 800 million weekly active users. Google Gemini reached 650 million monthly users. EMarketer projects US AI search advertising spending will surge from $1.1 billion in 2025 to $26 billion by 2029.
The current landscape (and what we’re learning from it)
ChatGPT: Just announced, still testing
OpenAI’s announcement came on January 16, 2026. Ads will appear at the bottom of responses in the free tier and the new ChatGPT Go subscription. Plus, Pro, and Enterprise subscribers won’t see ads. The company emphasises that ads will be clearly labelled and will not influence ChatGPT’s responses.
What we don’t know: pricing models, targeting capabilities, creative specifications, or how advertisers will actually access the platform.
What we’re watching: how users react. Will clearly labelled ads maintain trust, or will the perception of bias damage the platform’s credibility?
Perplexity: First mover, early signals
Perplexity launched ads in November 2024, becoming the first major LLM to integrate advertising. The format includes sponsored follow-up questions and paid media positioned alongside answers. Early advertisers include Indeed, Whole Foods Market, and PMG.
The pricing: £30-60 CPM (roughly $40-80 per thousand impressions). That’s significantly higher than standard display advertising, suggesting platforms are betting on high engagement rates.
Our observation: Perplexity’s model emphasises transparency. Ads are clearly marked as “sponsored,” and answers to sponsored questions are still generated by its technology, not written by brands.
Google Gemini: Leveraging existing infrastructure
Google Gemini already displays ads within its Search Generative Experience (SGE). With 650 million monthly users and deep integration into Google’s advertising ecosystem, Gemini has advantages in scale and advertiser familiarity.
What’s unclear: how these AI-integrated ads perform compared to traditional search ads.
We’re in month zero of a major shift. The platforms announcing ads doesn’t mean the playbook exists. It means we need to start paying very close attention.
Why we think LLM ads will work differently
Higher in the funnel (probably)
Google Search ads work because users arrive with clear intent: they’re comparing options, checking prices, ready to buy. LLMs serve a different purpose. People turn to ChatGPT or Perplexity to explore, learn, and get recommendations before they’ve narrowed down choices.
Our hypothesis: LLM ads will influence awareness and consideration, not immediate conversions. They’re discovery-phase advertising, less “Buy Now, 20% Off” and more “Here’s why this approach works for your use case.”
What we don’t know: whether platforms will develop lower-funnel ad formats over time.
Trust will make or break this
Every platform emphasises the same principle: ads won’t influence AI-generated answers. OpenAI, Perplexity, and others stress that responses remain objective and unbiased, with ads clearly separated and labelled.
Why it matters: users trust AI answers significantly more than traditional search results. According to research from Microsoft, visitors from LLMs convert to signups at 1.66%, compared to 0.15% from search engines. That trust is fragile. If users perceive that ads are manipulating responses, they’ll leave.
The risk: if trust erodes, the entire model collapses. Platforms are walking a tightrope between monetisation and credibility.
Lower ad load, higher stakes
Early observations: Perplexity shows 1-2 sponsored placements per session. Compare that to Google Search, which displays 4-7 sponsored results on a typical results page.
Our thinking: each LLM ad carries more weight. With fewer placements, creative quality and relevance matter significantly more.
Unknown: will ad load increase as platforms scale and revenue pressure grows?
Early data is promising (but limited)
Some initial signals suggest strong performance. Ahrefs research shows AI search traffic accounts for just 0.5% of total visits but generates 12.1% of signups.
Reality check: these are tiny sample sizes from early adopters. We don’t know if these conversion rates will hold as LLM advertising scales.
What we’re thinking about (and preparing to test)
I’m not claiming to have cracked LLM advertising. We’re preparing to experiment intelligently when access opens. Here’s what we’re focused on.
1. Building organic AI visibility first
Why this matters: You likely can’t advertise effectively for a brand the AI doesn’t already recognise. If ChatGPT or Perplexity never mentions your company organically, paid ads may fall flat.
What we’re doing: We’re auditing how WeAreBrain appears in AI-generated responses. We test queries like “best agencies for AI implementation in the Netherlands” and track whether we’re cited, how we’re described, and how frequently we appear.
The logic: If we’re invisible organically, paid ads alone won’t fix the visibility gap. Brands that already have strong AI citation rates will likely see better returns on paid ads, though we’re speculating, as this remains unproven.
2. Rethinking product and service descriptions
The shift: LLMs need context, not just feature lists. Generic descriptions like “WeAreBrain builds custom software solutions” don’t help AI systems understand when to recommend us.
Example of our thinking:
- Traditional: “Full-stack development agency”
- AI-optimised: “WeAreBrain builds custom software for startups and scale-ups needing MVP development, cloud migration, or AI integration, typically €50,000-€200,000 projects for teams of 10-100 people”
Why: More specificity theoretically improves LLM matching. If someone asks “who can build an MVP for a fintech startup with a €75,000 budget?” we want the AI to recognise we fit that profile.
Reality check: We’re speculating based on how LLMs process information. This approach could be completely wrong.
3. Preparing discovery-focused creative
Our hypothesis: LLM ads won’t work like Google text ads. The conversational, exploratory nature of these platforms suggests a different creative approach.
What that might mean:
- Less promotional: fewer “Buy Now, Limited Offer” messages
- More educational: “Here’s why this approach works for [specific use case]”
- Storytelling over sales copy: context and proof over features and pricing
Where we’re drawing inspiration: Podcast sponsorships, native content partnerships, formats that blend into the discovery experience rather than interrupting it.
We’re planning to test multiple creative angles when access opens. Storytelling versus direct value propositions. We genuinely don’t know what will perform. We’re preparing options.
4. Setting realistic expectations internally
What we’re telling clients: LLM ads probably won’t deliver last-click conversions immediately. These platforms serve users in exploration mode, not comparison or purchase mode. Expect ads to influence awareness and consideration, not replace performance channels like Google Search or paid social.
Why this matters: Overpromising on untested channels damages trust and leads to budget waste.
Budget implications: We’re recommending clients allocate small test budgets (think €2,000-€5,000 initially) to learn the platform dynamics before scaling.
5. Building attribution systems we can actually use
The challenge: Traditional UTM tracking may not capture in-chat interactions. If users never click through to a website but are influenced by an ad they saw in ChatGPT, how do we measure that impact?
What we’re exploring:
- Unique tracking codes specifically for AI platforms
- Separate analytics segments (ChatGPT already appends utm_source=chatgpt to outbound links)
- Post-conversion surveys: “How did you first hear about us?” with AI platforms as distinct options
What we don’t know: Will platforms provide conversion tracking APIs like Meta and Google? We’re preparing for both scenarios.
The big unknowns
These are the questions we’re asking, and nobody has answers yet:
- Pricing models: Will CPM remain standard, or will CPC and CPA options emerge as platforms mature?
- Targeting capabilities: How granular can targeting get without extensive behavioural tracking?
- Creative formats: Text only to start, but will image, video, or interactive ads follow?
- Minimum budgets: Can small businesses test affordably, or will high CPMs limit access to enterprise brands initially?
- Performance benchmarks: What constitutes a “good” CTR or conversion rate for LLM ads? We have no baseline yet.
I’m treating this like any emerging channel. Humble about what we don’t know, aggressive about learning when we can test.
Key takeaways
- LLM advertising is here. ChatGPT announced ads on January 16, 2026. Perplexity launched in November 2024.
- Nobody knows what works yet. Not agencies, not platforms, not advertisers.
- Prepare now by building organic AI visibility. You likely can’t advertise effectively for a brand the AI doesn’t recognise organically.
- Expect different results than Google Ads. LLM ads sit higher in the funnel, focusing on discovery and consideration.
- Stay humble and ready to test. The playbook is being written right now.
Want to talk through how your brand can prepare for LLM advertising? We’re tracking developments closely and planning our first tests when platform access opens. Let’s strategise together and make sure you’re positioned to move quickly when the opportunity arrives.
