The emerging LLM search advertising landscape: infographic

AI-driven large language model (LLM) search engines are going to completely change the way we find information, answer questions, search for products, and make purchase decisions. So it’s past time that we take a look at the emerging LLM search advertising landscape.

A visual overview of the LLM search advertising landscape

LLMs that matter: ChatGPT and more

As I write this, LLMs are growing super-fast, even if the LLM search advertising space is still incredibly young.

Here’s the ones that we can get data on:

• ChatGPT by OpenAI: 400M weekly users
• Ernie Bot by Baidu: 300M users (no details on MAU, DAU or WAU)
• Gemini by Google: 250M weekly users
• Llama by Meta: 200M weekly users
• Claude by Anthropic: 50M weekly users
• Perplexity AI: 15M weekly users

(Sources: BI, Reuters, Gemini estimate based on 106M app downloads plus 275M monthly visits plus integration into Google tools such as Gmail and Meet: 9M organizations, Llama estimate based on Mark Zuckerberg’s recent statement that Llama was serving 600M users monthly, Claude extrapolation from SimilarWeb, plus app usage, FT)

ChatGPT leads, hitting around 1% of global search volume by some measures. That seems small until you realize that with all the resources of Microsoft behind it Bing worked for a decade to capture share from Google and is still only under 4% of global share.

Gemini, Perplexity, and Copilot are also players in search, though with less usage. DeepSeek is a recent challenger.

And Llama by Meta might be a sleeper here: it’s not a standalone product or service like many of the others, and so much of its use might go unreported by analytics engines that detect traffic on the open web. Meta has embedded Llama in pretty much every significant product, so it could have a massive share that we’re just not seeing.

LLM search advertising versus traditional search engines

Of course, LLMs are much more than search engines. In fact, they’re vastly different than search engines. And that difference shows us what will change in an LLM search advertising platform.

Search engines are exactly that: focused on SEARCH. 

That’s actually a bad metaphor for what we wanted to accomplish in 1995 when we fired up Yahoo to look for something on the brand-spanking-new World Wide Web. And it’s a bad metaphor for what we want from Google and Baidu and Bing today.

We don’t want to search.

We want to FIND.

More than that, we want to know. To understand. Maybe, to make a decision. Perhaps to make a purchase.

Traditionally, a search engine provides links to sources — it retrieves and ranks web pages, leaving users to sift through results. In contrast, an LLM like ChatGPT delivers direct answers. It synthesizes information into a conversational, coherent response without requiring users to browse multiple sources. Search engines excel at finding specific web content, while LLMs focus on understanding context and generating responses tailored to user intent.

Of course, search engines have been changing over the years to offer more and more on-site zero-click answers, but LLMs take this to a new and vastly higher level.

Traffic from LLMs

We’ve seen the massive growth in LLM traffic here at Singular. As we reported in our latest Quarterly Trends Report, ChatGPT has become a top source of traffic for Singular customers without even offering an LLM search advertising platform.

The growth here has been astounding from 1 quarter to the next: 8,400%.

• Q3 2024: 1 Singular customer
• Q4 2024: 85 Singular customers

In Q3, just 1 Singular customer received traffic from ChatGPT. In Q4, that ballooned to 85, and the growth per customer increased massively as well.

And ChatGPT doesn’t even have an advertising platform yet: this is just pure organic traffic measured by 1 of the simplest and oldest of open web analytics solutions: UTM parameters: “utm_source=chatgpt.com” at the end of referrer links.

The ad networks are coming

LLMs are monetizing primarily via subscriptions right now, and that will continue for premium customers. But advertising always comes in to expand TAM and market share to those who won’t pay, and it is already partly here.

LLMs that have active ad platforms right now include:

• Gemini by Google
  Ads with answers
• Llama by Meta
  (Ads surrounding Llama)

LLMs that are planning or preparing their ad platforms include:

• ChatGPT
• Perplexity
• Claude

Expect every significant and open-to-the-public LLM to offer an ad network within the next 18 months or so. It’s essentially a requirement to capture value from all users, not just paying subscribers.

Ads for LLM search engines will be different

Ads will work differently on LLMs. And LLM search advertising platforms will be astly different than Google or Bing or Baidu.

How different we’ll find out over the next few years, but here’s some clues:

Targeting options:

• Untargeted brand ads
  Everyone needs soap. Big brands with broad targeting could adopt a mass media approach.
• Contextual targeting
  Someone searching for things to do in Tahiti might not have their flight yet..
• Behavioral targeting
  ChatGPT gets a pretty good sense of who you are as you use it, and that’s first-party data which could be used to create interesting audiences..
• Broad keyword matching
  Exact keywords won’t be necessary: LLMs understand language and can provide broad matching out of the box that is likely better than exact match keywords

Placement options:

• Answers are different than search results..
• Ads will be separate from answers..

Formats:

• Initial LLM ads have been text, and we’ll continue to see that
• We’ll also see some image+text ads, especially for sponsored product results
• We may see video ads playing side-by-side with the question-answer interface

We’ll probably also see a much lower ad load. Google, for example, often has dozens of ads per page. LLMs may not exactly have pages per se, and are likely to have single-digit numbers of ads per screen, at least initially.

Query uniqueness may go up as well: search engine queries have been getting longer and more conversational over time. Where we might once have typed “best pizza NYC” we’re more likely to ask “Where can I get the best pizza in New York City?”

That’s only accelerated with Alexa and Siri and Google Assistant, and now particularly with LLMs. In fact, I often use voice — even on my laptop — with ChatGPT to enter a phrase, sentence, or even paragraph of inquiry, which the LLM search engine then uses to find a very precise, specific, and contextualized answer.

Taste of the future

We’re only getting a taste of the future right now.

LLM search advertising measurement will come too, with programs for MMPs like Singular, and this will all get more sophisticated.

“This is only a taste of the future,” says Singular CMO Stephanie Pilon. “Perplexity has already set up ads on its LLM , and we’re seeing movement from Google’s Gemini as well. We expect LLMs like ChatGPT to be much more significant in marketers’ and advertisers’ plans in 2025.”