89% raised AI search budgets, 24% can prove a return: a measurement and ownership playbook for CPG brands

A familiar AI search problem at consumer brands is not technical at all. Someone was asked to "look into" ChatGPT a year ago. Now there is a budget, a few tactics in flight, a monthly slide, and no clear answer to the question a founder eventually asks: is this working?

A survey published on November 11, 2026 suggests that is the normal state of things. Peec AI, which sells AI search tracking, launched a quarterly series called Primary Sources. Its first edition pairs Peec's platform data, covered in our post on ChatGPT's citation shift, with a survey of 700 marketing decision-makers. This post is about the survey and what a packaged goods brand can take from it.

Who was asked

Sapio Research ran the survey for Peec AI in September 2026. Respondents were marketing decision-makers, manager level and above, with full or shared responsibility for AI search, at companies with 100 or more employees: 300 in the US, 200 in the UK and 200 in Germany. They included 270 owners and C-suite executives, 183 directors and VPs and 247 managers.

Two caveats before the numbers. The findings are self-reported: "can prove ROI" means the respondent says so. And the report itself notes that associations between practices and outcomes do not establish causation. Also, many CPG brands reading this have fewer than 100 employees, so the companies in the survey are larger than most of our readers. The patterns still translate, as we argue below.

Spending ran ahead of proof

The headline numbers, from the press release:

• 61% say increasing visibility and recommendations by AI platforms is a top priority.
• 60% checked what an AI assistant says about their own brand several times in the past month.
• 89% raised AI search spend in the past year, and 81% expect to raise it again.
• 24% can already prove a return.

The gap is wide even among the biggest spenders. Of companies that raised spend significantly, 64% can't show what it returned. And 54% of those who say they have reliable metrics they trust still can't show ROI.

The pressure to spend comes mostly from outside. The most-cited reasons to invest were competitors appearing in AI answers (43%) and customers using AI tools to research or buy (42%), ahead of the board or leadership (34%) and industry media and hype (28%).

Show the data

Market	Can prove ROI
US (300)	31%
UK (200)	19%
Germany (200)	18%
All 700	24%

Source: Peec AI, Primary Sources Q3 2026 (Sapio Research survey), Nov 11, 2026. Self-reported. Budgets rose at similar rates in all three markets: 87% US, 91% UK, 89% Germany.

US companies were further along. Budgets rose at about the same rate in all three markets, but 31% of US marketers can prove a return, against 19% in the UK and 18% in Germany. US companies were also the most likely to have built AI search into their wider marketing strategy: 38%, against 30% in the UK and 22% in Germany.

Why should a CPG founder care about proof, rather than just spending? Because the channel is worth measuring properly. Adobe reports that in September 2026, AI-referred traffic to US retail sites converted 90% better than non-AI traffic and generated 83% more revenue per visit. A channel that valuable deserves better than a guess.

What separates the teams that can show results

Size didn't. Mid-market and enterprise companies proved ROI at the same rate, 24%. What differed was how the work was organized.

WithWithout

Show the data

Practice	With	Without
Integrated AI search strategy	43%	23%
Dedicated full-time roles	35%	15%
Dedicated AI visibility platform	39%	24%

Source: Peec AI, Primary Sources Q3 2026, Nov 11, 2026. Self-reported association, not cause. Peec AI sells an AI visibility platform, so weigh the third row with that in mind.

Companies whose AI search strategy is integrated into their wider marketing were the most likely to prove ROI, at 43% against 23%. Dedicated full-time roles went with 35% against 15%, and the press release adds that teams with full-time roles were twice as likely to have metrics they trust (57% against 28%). Integrated companies also involved more of the business: 71% brought in data and analytics teams, 44% involved product (against 24%) and 37% involved engineering (against 20%).

Trusted measurement also kept teams from quitting too early. Among teams with reliable metrics they trust, 7% had stopped optimizing to be recommended by AI, against 31% of teams with some visibility they can't quantify. For content written for AI it was 8% against 25%.

The ownership problem

No single function owns AI search at most companies. The SEO team leads at 24% and executive leadership at 22%. Where more than one team is involved, 88% report coordination problems, most often conflicting priorities and metrics (39%) and waiting on other teams' capacity (35%).

The detail that stood out to us: comms or PR teams own AI search at only 3% of companies. And the tactic companies abandon most often is earning third-party mentions, dropped by 21%. That work usually belongs to PR. In Germany, 30% of marketers have stopped trying to earn third-party mentions, against 24% in the UK and 14% in the US.

Table: Share of companies using each AI search tactic

Tactic	Using it
Optimizing existing content	79%
Shaping how AI describes the brand	77%
Optimizing to be recommended by AI	75%
Creating content written for AI	74%
Making sites easier for AI systems to read	69%
Building a presence on Reddit, YouTube and LinkedIn	67%
Earning mentions in third-party media and reviews	57%

Source: Peec AI, Primary Sources Q3 2026. For every tactic, only 43% to 50% of the teams using it are very confident it works.

The playbook for a small CPG team

The survey describes companies with dedicated roles and data teams. Most founders reading this have a marketing lead, an ecommerce manager, an agency or two and a PR contact. Here is what we think the survey's lessons look like at that scale.

1. Name one owner. Not a committee. One person whose job description includes "what assistants say about our products," with an hour a week protected for it. At a small brand that is usually the ecommerce or growth lead, because they already own product data and retailer listings, which is where most of the facts assistants repeat come from.

2. Pick one steady measure and keep it. The teams that kept going were the ones that could see results. Write 20 to 40 questions a shopper would ask about your category and your products, ask them across the assistants your customers use each month, and record whether you are named, in what position and with which facts. Tip 12 covers setting it up, and why tracking a single AI "rank" misleads covers what not to measure. The tool matters less than the habit: a spreadsheet works, and so does a product-level score such as SearchPanel, which XENA Intelligence makes. What matters is that the questions don't change from month to month. One reason to track a list rather than a single answer: when Babel42 asked five assistants the same buying questions in early November, they put the same brand first only 27% of the time, even when all five named brands.

3. Put revenue next to it. Set up AI referral sources in your analytics as their own channel, as in tip 13, and report AI-referred sessions, conversion and revenue on the same page as the monthly answer check. If you have claimed your brand in ChatGPT's Business Hub, add its mentions and clicks; our Business Hub baseline post covers how to read them. This is the closest a small brand gets to the "integrated" pattern in the survey.

4. Bring PR in on purpose. If third-party coverage is the tactic most teams drop and PR almost never owns it, that is a gap you can fill cheaply. Give your PR contact the list of pages assistants cite for your category questions, and ask them to target those publications and reviewers. Tip 11 covers getting into cited sources.

5. Give each tactic a quarter. Platforms change fast. Peec's own data shows ChatGPT's citation mix moved within a week of a model change. That makes it tempting to switch tactics monthly. Don't. Decide what you are trying, write down what you expect to move, and judge it after a full quarter of the same measure.

6. Hold the budget until the baseline exists. If you can't yet say what your AI visibility was three months ago, the next dollar is better spent on measurement than on more tactics. Tip 44 applies the same logic to ChatGPT ads.

What we would not take from this survey

We would not read the "dedicated platform" row as proof that buying software causes ROI. It comes from a platform vendor, it is self-reported, and teams that buy a platform may simply be the ones already organized to use it. The same caution applies to every row in that chart. The survey's clearest message is about organization: one owner, a shared measure and the right teams involved. Those are free to adopt, and they are what a founder can control this quarter.