Report/ Primary Sources, Q3 2026

AI search is a priority. Commitment sets teams apart.

Customers are using AI to research what to buy. Competitors are showing up in the answers. Companies know they need to invest in AI search, but many are still working out how to organize the work and measure what it delivers.

Primary Sources combines a survey of 700 marketing decision-makers responsible for AI search with Peec AI’s analysis of platform behavior. See what’s driving investment, how teams are organizing the work, and what it takes to keep pace.

Key findings

Market signals are driving investment
Competitors appearing in AI answers (43%) and customers using AI tools to research or buy (42%) are the most cited pressures to invest, ahead of board or leadership pressure (34%) and industry media or hype (28%).
Commitment matters more than company size
Only 24% of companies say they can prove ROI. Mid-market and enterprise companies report the same rate. Organizational commitment shows clearer differences: reported ROI is higher among companies with an integrated strategy (43% versus 23%), dedicated full-time roles (35% versus 15%), and a dedicated AI visibility platform (39% versus 24%).
Platform changes demand a new approach
Rapid platform change is a barrier for 33% of companies. After ChatGPT made GPT-5.6 its default on August 8, sessions running multiple searches rose from 2% in July to 61% in September. Product-page citations gained ground as Reddit citations fell. Teams need ongoing monitoring to spot these shifts and adjust their content priorities.

Summary

Marketing teams plan by the quarter, while AI platforms change week to week on different schedules. The challenge is connecting shifts in searches and citations to decisions about where the team should focus next.

Spending more has become the default response. 89% of companies raised AI search spend in the past year and 81% plan to raise it again, yet three in four that raised it can’t show what it returned. This edition is about what money can and can’t do, and what the teams getting past that point have changed.

Inside the report

Chapter 2
Investment is ahead of proof
What companies measure today, which metrics they trust, how often they trace leads or revenue to AI search, and when they expect to prove ROI.
Chapter 6
How teams judge their tactics
Which tactics companies use, which they abandon, and how trusted metrics relate to whether teams stick with their work.
Chapter 7
AI search changes faster than teams can plan
How ChatGPT’s search behavior changed during Q3, including its use of site: searches, how its cited sources shifted, and how those patterns differ across platforms.

Product pages roughly doubled their citation volume in ChatGPT after the model switch, while Reddit citations fell. The pages ecommerce teams have ignored in favor of better-performing content are now being read by ChatGPT.
**Malte Landwehr**CPO & CMO, Peec AI

Contributors

The marketers and search experts outside Peec AI who shared their view of the findings.

Methodology

Where the data comes from

Two primary sources: a survey we commissioned, and our own analysis of AI platform behavior.

Decision-maker survey

Conducted for Peec AI by Sapio Research. Respondents hold full or shared responsibility for AI search. Findings are self-reported; associations between practices and outcomes do not establish causation.

Respondents: 700 marketing decision-makers
Markets: US (300), UK (200), Germany (200)
Seniority: 270 owners and C-suite, 183 directors and VPs, 247 managers
Company size: 100 to 999 employees (mid-market) and 1,000+ (enterprise)
Topics: Strategy, investment, staffing, tools, tactics, measurement

Peec AI platform analysis

Fan-out queries and cited sources. Fan-out analysis covers ChatGPT’s default free experience (GPT-5.5, then GPT-5.6 from August 8); source and citation analysis covers five platforms.

Period: July to September 2026, weekly trends indexed to a Q2 2026 baseline
Platforms: ChatGPT, Claude, Gemini, AI Overviews, AI Mode (Perplexity excluded for data quality)
Measures: Fan-outs per session, query modifiers, sources per answer, answer length, citations by source type and platform