The rise of advertising in AI search: what marketers should expect in 2027

Ads are already inside ChatGPT, Google AI Mode and other AI surfaces. Here is what the 2026 data shows, what forecasters expect for 2027, and how to plan budgets, landing pages and measurement.

Key takeaways

  • Ads inside AI answers are live now. Promptwatch Data found ads in an average of 20.1% of ChatGPT web-search responses between May 20 and August 17, 2026, with a peak day of 43.9%.
  • Forecasters expect conversational ads to become a multibillion-dollar channel in 2027, but the numbers come from vendors and analysts with different incentives. Treat them as direction, not gospel.
  • Queries in AI Mode are longer and more exploratory, so ad targeting is shifting from keywords toward situations and needs.
  • Agents will sit on both sides of the transaction: brand agents answering clicks and buyer agents researching for people. Pricing, specs and structured data become part of your ad strategy.
  • You need two measurement tracks: paid performance, and organic AI visibility (who gets cited and mentioned next to or instead of your ad).

Where AI search advertising stands going into 2027

A year ago, "advertising in AI search" was a slide in a future-trends deck. Now it is a line in actual media plans. ChatGPT served no ads at all until May 27, 2026. By the middle of the summer, ads were showing up in roughly a third of the responses that triggered a web search.

That speed is the story. Most of the commentary I read for this guide agrees on the direction, even when the details differ. Adswerve expects conversational advertising to create "new touchpoints inside the AI interfaces where consumers increasingly start their research." Forrester, as cited by ad-platform vendor Lapis, predicted on September 17, 2026 that conversational ads would become a multibillion-dollar channel in 2027.

A word of caution before the numbers. Several of the figures below come from vendor blogs and secondary summaries, not from audited disclosures. I've labeled the source each time. Where I could check against first-party data, I did.

The 2026 scoreboard

This is the set of signals Lapis compiled as the starting point for 2027 planning. They are reported by the companies named, relayed through Lapis, so verify before putting them in a board deck.

SignalReported numberAttributed to
ChatGPT weekly users1.2 billion+OpenAI, Oct 21, 2026
ChatGPT Ads availability89 countriesOpenAI, Oct 21, 2026
Advertisers that have run ChatGPT ads100,000+OpenAI, Oct 21, 2026
ChatGPT Ads run rate$1B annualized in under 200 daysOpenAI, Aug 31, 2026
Google AI Mode monthly users1 billion+Alphabet Q2 2026 earnings call
Advertisers using Google AI Max500,000+Alphabet Q2 2026 earnings call
US AI search ad spend forecast$2.08B (2026) to $25.93B (2029)eMarketer, Jan 2026
Marketers raising AI assistant investment in 2027Net 75%Kantar, Oct 2026

There is a skeptical read too. Lapis points to a report that OpenAI is likely to miss the $2.4 billion ad revenue it projected to investors for 2026. The same piece cites New Street Research estimating ChatGPT advertisers tripled from about 20,000 to roughly 60,000 between late August and early October, with clicks near $5 versus $30 to $50 for comparable commercial software ads on Google.

My take: cheap clicks and a missed revenue target are two sides of the same fact. Inventory is underpriced and under-supplied by advertisers. That usually does not last, which is why 2027 is the year to test while it is inexpensive.

What first-party data says about ChatGPT ads

Most AI ad commentary is opinion. Promptwatch publishes actual observations. In its ChatGPT Ads Over Time report, covering May 20 to August 17, 2026, only responses where ChatGPT searched the web, returned at least one citation and completed ad extraction were counted. The findings:

  • Ads appeared in an average of 20.1% of those responses over the 90 days.
  • No ads at all until May 27, when they showed in 1.69% of responses.
  • The rate jumped past 30% on July 1 and mostly stayed between the low 20s and high 30s afterward, with a peak day of 43.9%.
  • The past 7 days averaged 32.4%, and the past 30 days averaged 29.6%.
  • By prompt type, ad-bearing responses over the past 7 days were 73.3% organic prompts, 18.0% brand-specific and 8.6% competitor comparison.

Two cautions from the report itself. That last split reflects the mix of prompts being monitored, so it shows where ads appeared, not how likely each prompt type is to trigger one. And a single run can land on an outlier day, so check ad visibility across several days before drawing conclusions.

Still, one practical point stands out. Organic, non-branded prompts carried most of the ads in the monitored set. That is where buyers are researching without a brand in mind, and where you are most likely to compete with ads for attention.

For shopping surfaces, Promptwatch's ChatGPT Shopping usage data shows product cards attach to only a low single-digit percentage of web-search responses overall, with far higher rates on commercial and transactional prompts. The trigger rate moves in steps, which the report reads as OpenAI actively tuning when product cards appear. Expect the same tuning with ads.

How AI search changes what you are bidding on

Keywords were built for short queries. AI queries are not short. According to an SEJ write-up of Google's Ads Decoded episode (October 28, 2026), Ginny Marvin said queries in AI Mode are three times longer than traditional searches, and discovery is moving up the funnel as people use conversational search for brainstorming well before purchase.

The same report cautions against over-reading it. Longer questions give advertisers more information about someone's needs, but the figure by itself does not show stronger buying intent or better conversion rates. It also noted that more than one in six AI Mode searches in the US are multimodal, which is narrower than the "one in six searches" phrasing in some posts.

For paid teams, SEJ's practical suggestions are sensible:

  • Mine customer questions, sales calls and search terms for problems people describe before they know your product name, then test ad messaging against those needs.
  • Put price, compatibility, setup and limitations where visitors can find them fast.
  • Link reviews, demos and customer stories from landing pages, tied to a known concern.
  • Check whether research visits turn into later purchases over the normal buying cycle, not just impressions or CTR.

That last point matters more in 2027. If an AI answer does half the persuading before a click, last-click reports will undercount ads and organic content alike.

Agents on both sides of the ad

Two separate things get called "agentic advertising," and mixing them up causes bad planning.

Agents that buy and run your ads

Adswerve expects specialized agents handling campaign adjustments, fraud detection, real-time audience updates and creative optimization, with human oversight. Lapis cites IAB Europe research that 58% of ad executives expect agentic buying to reach operational use or scale within a year. That is a vendor-relayed number, but it matches what platform roadmaps suggest: Google's AI Max, Meta's business agents and similar products are moving manual levers into software.

The constraint, as Adswerve puts it, is that agents only perform as well as the data behind them. If your conversion data is late, messy or ungoverned, an agent will optimize confidently in the wrong direction.

Agents that read your ads and shop for people

The other side is buyer-side agents. Adswerve argues marketers will need two parallel strategies: one that earns trust with people, and one that makes products "discoverable, credible, and transactable for the machines acting on those people's behalf." Fourfront's 2027 SEO predictions add a useful corrective: agents will research for buyers but will not close the deal, and hiding pricing behind "contact us" will leave you out of the conversation.

I agree with both. A buyer agent comparing five vendors cannot compare the one that will not publish a price. Treat pricing pages, specs, FAQs and product schema as ad-adjacent assets. They decide whether you appear in the shortlist before any bid is placed.

What to budget

Lapis suggests, for a performance budget, roughly 12% in conversational and AI answer ads, 5% in agent formats, 3% in AI app networks and 5% in experiments and measurement, keeping 75% in core search and social. That is one vendor's recommendation, and Lapis sells AI ad software, so adjust for your own category. But the shape is reasonable: a minority slice, explicitly labeled as testing.

A simple way to run that test:

  1. Pick 20 to 50 non-branded prompts where your buyers research.
  2. Record who wins the ad slot and who gets cited organically, across several days.
  3. Launch a small campaign against those themes with distinct landing pages.
  4. Compare assisted conversions over a full buying cycle against your core channels.
  5. Keep what beats your blended cost, cut what does not.

Measuring ads and organic visibility together

This is the part most teams skip. An AI answer can show your ad, cite your competitor as the source, and mention your brand in the prose, all in one response. Paid dashboards see only the first.

Tools I would look at, by job:

JobWhat to look forExample tools
Track which advertisers appear on your promptsAd copy, claims, landing pages per promptPromptwatch Ads Radar
Track organic mentions and citationsMulti-model, multi-country prompt monitoringPromptwatch, Profound, Peec AI, Otterly.AI
See whether AI bots can read your pagesCrawler logs, site auditsPromptwatch, DarkVisitors
Attribute AI-referred trafficReferrer plus UTM handlingPromptwatch visitor analytics, Cometly
Plan content from AI query gapsQuery fan-outs, content gap analysisPromptwatch, AirOps

Promptwatch's Ads Radar shows which advertisers bid on your prompts, their ad copy and claims, and the landing pages they send people to, viewed by ad or by prompt. It sits next to citation tracking, crawler logs and visitor analytics, so you can see the paid and organic sides of the same prompt in one place. Ads Radar is on the Business plan ($579/mo), every agency plan and Enterprise; confirm current details at promptwatch.com/pricing.

Favicon of Promptwatch

Promptwatch

Track and optimize your brand's visibility in AI search engines
View more
Screenshot of Promptwatch website

One attribution wrinkle: ChatGPT stopped sending a referrer on October 8, 2026, so some AI visits now look like direct traffic. Tagging your ad and link URLs with utm_source=chatgpt.com, and counting those in analytics, is a cheap fix. The Academy guide on attributing AI traffic walks through it.

If you are comparing monitoring tools more broadly, the GEO software directory at bestgeosoftware.com lists dozens, and many are priced for small teams.

Favicon of Profound

Profound

Track and optimize your brand's visibility across AI search engines
View more
Screenshot of Profound website
Favicon of Peec AI

Peec AI

Multi-language AI visibility tracking
View more
Screenshot of Peec AI website
Favicon of Otterly.AI

Otterly.AI

Affordable AI visibility monitoring
View more
Screenshot of Otterly.AI website

Organic still decides who wins

Ads are a layer on top of an answer that is built from cited sources. If the cited sources do not include you, the ad has to do all the work, and at that point you are paying for visibility you could partly earn.

Promptwatch's data on which content types ChatGPT cites most found product pages were the most cited type at 22.4% of citations (Sep 28 to Oct 4, 2026), followed by landing pages and listicles. Its half-life study found landing pages and docs stayed cited longest, while news lost half its cited pages within 36 days. So the pages you build for ad landing and comparison are also the pages AI is most likely to keep citing. Build them once, make them honest and complete, and they work in both channels.

It is also worth knowing that ChatGPT has pulled back from social sources. Promptwatch's October 2026 social citation data showed ChatGPT drew 0.46% of citations from social platforms versus 12.2% for Google AI Overviews. Your channel mix for earning mentions should differ by engine.

A 2027 checklist

  • Put real prices, plans and limits on public pages. Hidden pricing excludes you from agent comparisons.
  • Add complete Product schema, accurate availability and a clean merchant feed if you sell goods.
  • Audit landing pages against the questions customers ask before buying, not just the keyword.
  • Check your server logs or a crawler analytics tool to confirm AI bots can fetch your key pages.
  • Reserve a small, labeled test budget for conversational and AI answer ads.
  • Fix your first-party conversion data before handing campaigns to agents.
  • Review both paid and organic AI visibility weekly, across several days, since single runs mislead.
  • Measure over the full purchase window. Impressions and CTR will mislead you.

What I would not do

I would not shift core search budget on the strength of a forecast. eMarketer's $25.93 billion by 2029 is a projection, and the $1 billion run rate comes from the seller of the inventory. I also would not trust any "best partner" claim from a vendor describing itself, and I would not assume that longer queries mean better leads.

What I would do is spend a small, deliberate slice learning how these ads behave in your category while clicks are cheap, and fix the pages and data that every channel, human or machine, depends on. The pricing will probably rise as more of those 100,000 advertisers learn what works. Starting in Q1 2027 beats starting after that.

A free way to begin: run Promptwatch's free AI visibility report to see how ChatGPT, Gemini and Perplexity currently talk about your brand, then decide where ads fit around the gaps.

Share:

© 2026 AI Search Tools · Best AI search tools and platforms · RSS

AI Search Tools is an affiliate review site. When you click links to vendors or buy through links on our site, we may earn an affiliate commission at no extra cost to you.

AI Search Tools is a review website based on user reviews on Reddit and G2, and on publicly available information. We keep everything as up to date as possible, but pricing and features can change. Always confirm the details with the vendor before purchasing.