For most of this year, the prevailing wisdom has been that new AI channels are where the growth is and Google search is the safe, boring line on the plan that nobody needs to question.
This week turned both halves of that on their head.
The first independent study of ChatGPT ads found Google search visits earned almost three times as much. At the same time, media buyers told Digiday they are asking Google to justify a slice of its budget for the first time in years. The UK regulator wants Google to start reporting AI search clicks by 3 December. And Amazon launched an AI agent that will plan and buy your campaigns, while also grading its own homework.
The thesis running through all of it: no channel gets a free pass any more. Old or new, AI-powered or not, every pound now has to prove what it returns, and the brands that win will be the ones that build their own evidence instead of trusting the platform’s.
Here is what mattered, why it mattered, and what to do with it on Monday.
Board Analysis, The Digital Maze’s weekly marketing briefing, goes deeper on all of this every Friday. Subscribe here.
The Big Story: ChatGPT Ads Finally Have A Benchmark, And It Isn’t Kind
Last week, OpenAI’s ad business passed a $1 billion run rate without publishing a single performance benchmark. This week, academics did it for them. Search Engine Journal reports that “In U.S. traffic, a visit from a ChatGPT ad brought in $0.89 in revenue, against $2.45 for a visit from a paid Google search ad, according to a working paper by Maximilian Kaiser and Christian Schulze.” Conversion rates told the same story: “ChatGPT ad visits converted at 0.37 percent in U.S. traffic versus 1.17 percent with paid search.” And of the advertisers studied, “A total of 73 had zero sales generated by their ChatGPT ads (across 18,726 visits between them).”
This is the thesis in one story. A new channel arrived with huge scale and no proof, and the first proof says it’s well behind the channel it was meant to disrupt. That doesn’t make ChatGPT ads worthless. The data is U.S. only, ends on 22 August, and one advertiser drove 66 percent of all purchases. But it does mean the burden of proof sits with the new channel, and you should price it accordingly. The paper’s “parity price” is the most useful idea in it: the authors value a ChatGPT ad click at about $0.26, while “OpenAI’s ads basics page recommends starting CPC campaigns at a maximum bid of $3 to $5.”
Why this matters:
- Of the 87 advertisers who bought both, 84 earned less per ChatGPT ad visit than per search visit. Don’t fund a ChatGPT test by cutting search that already pays.
- Work out your own parity price: divide ChatGPT revenue per visit by search revenue per visit, then multiply by your search cost per click.
- OpenAI’s own FAQ says ChatGPT Ads “does not yet have performance benchmarks across advertisers, industries, or campaign types”, so your test is the benchmark. Filter out ad-review crawler visits before you judge it.
Read more: Google Search Ad Visits Out-Earned ChatGPT Ads In U.S., Search Engine Journal, 2 November 2026
Search & Intelligence: The UK Regulator Wants Proof Of What AI Search Sends You
Right now, you can see how often Google’s AI features show your site, but not how many people click. Search Engine Journal explains that the Search Console Generative AI report counts impressions, “However, the page doesn’t list clicks, CTR, or queries.” That may be about to change in the UK. The Competition and Markets Authority “lists clicks and click-through rate among the AI search metrics Google must provide”, and “The CMA gives Google until Dec. 3 to comply.” Asked about it, Google’s Gary Illyes said: “The team is super interested in giving more data, data that makes sense, but we don’t have anything to announce yet.”
For UK marketers, this is the proof thesis applied to organic search. For two years we’ve argued about whether AI Overviews and AI Mode cost clicks without being able to measure it properly. The CMA also wants publishers to be able to tell whether content appeared in an AI feature or an organic result. If that data lands, a lot of SEO reporting will need rebuilding around it, so set your board’s expectations now and plan a re-baseline for December.
Why this matters:
- Today’s AI report shows impressions only, so any “AI clicks” figure you’ve seen is an estimate.
- The CMA wants AI and organic appearances reported separately, which could change how you judge SEO performance.
- Page-level controls for AI search features are due three months after the 3 December deadline.
Read more: Google Says No AI Click Reporting News Yet, Ahead Of UK Deadline, Search Engine Journal, 2 November 2026
Also worth your time this week:
Paid Media & Performance: Google Search Has To Earn Its Budget Now
Digiday spoke to 18 media buyers and found clients “putting Google through the kind of scrutiny usually saved for newer channels.” The numbers back it up: “Search ad revenue grew 12% in the third quarter, Alphabet’s slowest pace this year, and executives blamed part of the slowdown on AI Mode”. One buyer put the size of the question plainly: “The amount being questioned is generally around 10% to 20% of the Google budget when that conversation comes up.” And with “ads in AI Mode are already expanding to new markets in Brazil, Mexico, and the United Kingdom”, the search results page your ads sit on is changing too.
Put this next to the Big Story and the thesis is clear. ChatGPT ads have to prove they work, and so does Google. The answer isn’t to pull search because the headlines say so; the same study shows search still out-earns the new kid. It’s to test incrementality properly, so you know which part of your Google spend creates sales and which part just collects credit for them. As one buyer said, “The incremental 10% to 20% is becoming less automatic.”
Why this matters:
- Q3 search ad revenue grew 12%, Alphabet’s slowest pace this year, as AI Mode changes what a click is worth.
- Expect your board to ask which 10% to 20% of search spend is truly incremental. Have the test results ready.
- AI Mode ads are coming to the UK, so plan how you’ll measure them separately from standard search.
Read more: Google search loses its free pass with buyers, Digiday, 2 November 2026
Also worth your time this week:
Web Development & Design: Your PageSpeed Score Just Changed. Your Website Didn’t.
PageSpeed Insights now runs on a new version of Lighthouse, Google’s testing tool. Search Engine Journal reports that “PageSpeed Insights is now using Lighthouse 13.6.0, which was released Oct. 27”, and that “A page may receive a different PageSpeed Insights score without anything on it changing, due to the change in the test.” Google’s Barry Pollard says the simulated loading metrics should now be more accurate and generally score higher, especially for JavaScript-heavy sites.
This is the proof thesis in technical form. A higher score after this update might be your developer’s work, or it might be the new test. If you report speed to a board or a client, you could accidentally claim a win you didn’t earn, or miss a real problem hiding behind a kinder score. SEJ’s advice is simple: “Before reporting an improvement, check that both reports name the same Lighthouse version, and compare the field data.” Field data comes from real Chrome users, and it’s what actually reflects your customers’ experience.
Why this matters:
- Lab scores from before and after 27 October aren’t measured the same way, so trend charts may show a false jump.
- JavaScript-heavy sites are the most likely to see their scores move.
- Lighthouse 13.6 also checks for an ard.json file, a sign that readiness for AI agents is becoming a standard audit.
Read more: Google Lighthouse Update Changes How PageSpeed Insights Simulates Loading, Search Engine Journal, 2 November 2026
Also worth your time this week:
Brand & Connection: John Lewis Proves Restraint Still Sells
John Lewis launched its 20th Christmas ad this week, and the brief came from proof, not a hunch. The Drum reports that “The brief was born out of John Lewis research showing that around half its customers lack confidence when it comes to giving gifts. Many end up playing it safe and tucking money into a card.” The result is Dame Kristin Scott Thomas, a talking toy dog called Pickles and the line “Don’t give up on giving.” Asked why the ad isn’t packed with more product, CMO Anna Braithwaite said: “We are a brand of restraint.”
Brand building has to prove itself too, and the evidence this week was on its side. A new IPA and Bauer Media study found “Campaigns combining audio, OOH and publishing report 33% more brand effects and 8% more business effects than campaigns using none of these channels.” John Lewis’s approach shows what that looks like in practice: a real customer insight, a product that stays in the background, and a year-round gifting campaign, including a vodcast and an in-store creator studio, carrying the idea well beyond one TV spot.
Why this matters:
- The strongest brief starts with customer research. John Lewis found around half its customers lack gifting confidence.
- “A brand of restraint” is a choice. Not every ad needs every product in shot to sell.
- The ad sits inside a year-round campaign, so the idea keeps working after the launch spot.
Read more: John Lewis Christmas ad: Dame Kristin Scott Thomas befriends talking toy dog Pickles, The Drum, 3 November 2026
Also worth your time this week:
AI & Automation: Amazon Wants An AI Agent To Run Your Ads. Who Checks Its Work?
Amazon has put AI at the centre of its ad business. Digiday reports that with Amazon Ads Agent, “Advertisers can now use natural-language instructions to plan campaigns, generate creative, forecast audience reach, adjust budgets, and analyze performance.” Amazon is confident it will pay off: “We expect year-over-year advertising revenue to grow in the high teens in fourth quarter 2026,” reads its filing. It also claims beta testing showed “up to 67% higher return on long-term ad spend and 29% lower cost per new-to-brand customer than manually configured full-funnel campaigns.”
Here’s where the thesis bites hardest. Those performance claims come from the same company that runs the agent, sells the inventory and holds most of the purchase data. Digiday notes that in Full-Funnel Campaigns “advertisers approve creative but cannot select individual channels or specific audiences.” As one adviser put it, “Amazon is open where openness brings it more of the budget and closed where its advantage sits.” Automation can save your team real time. Just make sure you still have an independent way to check what it’s doing with your money.
Why this matters:
- Amazon expects ad revenue to grow in the high teens in Q4, so expect a hard push on Ads Agent before Christmas.
- The 67% return claim is Amazon’s own beta data. Ask for an incrementality test using your own customer data.
- Before handing over control, ask what you can inspect, what you can verify independently, and whether fees stay visible.
Read more: Amazon’s Ads nears $20 billion in Q3 as it bets on agentic AI to fuel further growth, Digiday, 30 October 2026
Also worth your time this week:
From Our Blog: Charities, Google Will Pay For Your Search Ads
While everyone else argues about what a click is worth, eligible UK charities can get theirs paid for. In Are You Eligible for a Google Ad Grant?, Emlyn Dennis explains how Google Ad Grants give eligible charities and nonprofits up to $10,000 (approx £7,500) per month in free Search advertising, who qualifies, and the compliance rules, like keeping account-wide CTR above 5% each month, that catch charities out. Our free eligibility checker tells you in under 2 minutes whether your charity qualifies.
Case Study of the Week: +81% Conversions In The Quietest Season Of The Year
Buckley Loft Conversions builds high-quality loft spaces, in an industry with a predictable slow season. Their previous marketing failed to generate enough enquiries during the quiet months, leaving a gap in their pipeline.
We built an integrated paid media strategy to capture existing demand and create new interest. We restructured their Google Ads campaigns around high-intent loft conversion keywords, retargeted visitors who showed interest but didn’t convert, and launched Facebook and Instagram campaigns aimed at homeowners in their service areas, paired with strong visuals and seasonal offers.
- +81% increase in conversions
- -23% decrease in CPA
It’s the same lesson as this week’s news. Budget doesn’t earn its place by default, in a quiet season or in a new channel. It earns it by being aimed at real intent and measured against what it actually delivers.
Read the full Buckley Loft Conversions case study.
Bringing It Together
Every story this week asked the same question: can you prove it? ChatGPT ads finally met an independent benchmark and fell short. Google search, the line nobody used to question, now has to justify its share of the budget. The UK regulator wants Google to show publishers what AI search actually sends them. Even your PageSpeed score needs checking before you celebrate it. And Amazon’s new agent arrives with impressive numbers that only Amazon can verify.
The brands that win won’t be the ones that pick the right channel based on hype, or stick with the old one out of habit. They’ll be the ones that build their own evidence: parity prices, incrementality tests, consistent reporting and real customer insight, like the research behind John Lewis’s ad.
Three Takeaways for the Week Ahead
- Price ChatGPT ads on evidence, not enthusiasm. One study found ChatGPT ad visits earned $0.89 against $2.45 for paid search, and valued a ChatGPT click at about $0.26 while OpenAI suggests starting bids of $3 to $5. Work out your own parity price before you scale, and check your ChatGPT Ads settings now, because creative enhancements that can rewrite your ads and landing pages switched on by default on 30 October.
- Make every channel prove its incremental return. One buyer says around 10% to 20% of Google budgets is now being questioned, and Amazon’s new Ads Agent controls both the buying and much of the measurement. Use your own customer data and incrementality tests to judge platform claims, rather than the platform’s reporting alone.
- Reset your reporting before December. PageSpeed Insights scores have shifted with Lighthouse 13.6.0, and the CMA expects Google to report AI search clicks and CTR from 3 December. Note the Lighthouse version on every speed report, and agree with your board now how AI search will be measured once the new data arrives.
Want help putting any of this into action?
Whether it’s testing ChatGPT ads properly, proving what your Google budget really returns, or getting your reporting ready for AI search data, our team can help you turn this week’s changes into a plan. We combine SEO, PPC and web development so your marketing works as one joined-up strategy, with every decision tied back to revenue.
Book a call with The Digital Maze.
Want this in your inbox every Friday? Board Analysis is our weekly briefing for UK marketing leaders: genuinely useful marketing updates, delivered weekly.
