Marketing has spent the last decade getting more measurable. Every click tracked, every conversion attributed, every pound accounted for in a dashboard. The assumption has been simple: if you can see it in the data, it’s real.
This week, that assumption took a beating.
Apple quietly started blocking ad tech trackers across every iPhone browser. Agencies told Digiday that ChatGPT reported around 100 ad clicks where their own tracking found 20. Amazon argued that around a fifth of a brand’s value arrives after the attribution window has closed. Google told sites to fact-check their AI content and called fake expert profiles “a form of deception”. And Les Binet and Kantar found that 75% of marketers can’t estimate what proportion of their promotions are profitable.
The single thesis running through this week’s news is that the numbers marketers rely on are getting less reliable, and the advantage is shifting to brands that can prove what’s real. That means real tracking you control, real experts on your content and real profit, not just activity.
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: Apple Can Now Switch Off Your Tracking Whenever It Likes
Digiday reports that Apple’s iOS 27 updates, “released in mid-September,” mean Safari “would stop talking to those companies altogether.” The first companies blocked were The Trade Desk, LiveRamp, ID5, Permutive and Experian’s Audigent. Digiday has seen “results from a test of more than 70 ad tech domains that found only that first handful blocked.” The bigger worry is what comes next. Apple has written new code into WebKit that “will let it add companies whenever it likes without waiting for a software update.” As Rowena Lam of IAB Tech Lab put it: “What this will allow Apple to do is make this block list dynamic.”
The reach goes well beyond Safari. “Apple requires every browser on the iPhone and iPad to run its own engine, WebKit, meaning Chrome, Firefox and the rest are caught up too.” And iPhone users on Safari “are among the most valuable readers on the web.” Most of the industry would only discuss it off the record, which tells you how much is at stake. For UK brands, the takeaway is uncomfortable but simple: if your targeting and measurement depend on third-party trackers, some of your most valuable audiences can drop out of your data overnight, without a release note to warn you.
Why this matters:
- Apple can now add companies to the block “whenever it likes”, so your reporting can change without any announcement.
- Every iPhone and iPad browser runs on WebKit, so switching browsers doesn’t get your audience back.
- ID vendors are testing workarounds “including moving their pixels server-side”, which is where your own tracking needs to head.
Read more: Apple’s secret blocklist is the talk of Advertising Week New York (mostly off the record), Digiday, 7 October 2026
Search & Intelligence: Google Wants Proof A Human Checked It. And A Real Human Wrote It.
Search Engine Journal reports that Google revised its guidance on using generative AI content on 1 October. The new wording warns that “generative AI outputs may contain inaccuracies (also known as hallucinations). It is critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing.” It also changed “This includes metadata” to “This review also applies to metadata,” which brings titles, meta descriptions, structured data and image alt text into scope. A day later, SEJ spotted a new paragraph on Google’s people-first content page: fabricating creator profiles “is a form of deception,” and “Any form of deception makes a page untrustworthy to both users and our automated quality systems, and is a signal of a low-quality page.”
Put the two together and Google’s position is clear. AI is fine as a tool, but a human has to check the output, and the expertise you claim has to be real. That’s a direct challenge to the content shortcuts many brands have taken, from AI-written meta descriptions nobody reads before they go live to “expert” bylines with stock headshots. The same week, SEJ reported that Google’s AI contribution pilot pays some publishers “less than 0.1% of their advertising revenue.” The message to content owners is that Google will hold you to high standards, while the rewards for feeding its AI stay small. Trust has to be earned on your own site, by your own people.
Why this matters:
- Google now says to fact-check “all AI-generated content” before publishing, so build a human review step into every workflow.
- The review “also applies to metadata”, so AI-written titles, descriptions and alt text need checking too.
- Fabricated author profiles are now named as “a signal of a low-quality page”, so put real experts on your content.
Read more: Google Tells Sites To Fact-Check AI Content Before Publishing, Search Engine Journal, 1 October 2026
Also worth your time this week:
Paid Media & Performance: ChatGPT Ads Can’t Prove They Work Yet. Budgets Know It.
Digiday reports that “Accuracast ran a U.K. lead-gen campaign on ChatGPT alone and watched the form fills come in, but two weeks later the platform still showed zero conversions.” The same agency “found ChatGPT reporting around 100 clicks where its own tracking picked up 20,” and where conversions do register, “the agency sees them arrive 24 to 36 hours late, against a few hours on Google. Reporting carries no demographic or prompt data.” The budget impact is plain: one exec said they have clients “spending $10 million a month on Google,” but put “less than $100,000 a month into ChatGPT despite wanting to spend more.” Accuracast said the unanswered questions stop it moving clients “from test budgets between £10,000 to £20,000 up to £100,000.”
OpenAI knows this is the problem. This week it announced new conversion integrations, attribution partners and a visual ad format that Search Engine Journal says “will begin later this month with an initial group of U.S. advertisers.” But Amazon’s pitch this week shows measurement is a fight on every platform. Digiday reports Amazon’s claim that “around a fifth of a brand’s value on Amazon arrives after the attribution window, typically seven to 30 days, has closed,” a figure drawn from its own study of more than 2 million campaigns. Every platform is asking you to trust its numbers. The brands that come out ahead will be the ones that check those numbers against their own data before the budget moves.
Why this matters:
- A UK lead-gen campaign showed zero conversions in ChatGPT two weeks on, so keep your own conversion tracking in place.
- ChatGPT reported around 100 clicks against 20 in the agency’s tracking, so never take platform click data at face value.
- Agencies are holding test budgets at £10,000 to £20,000, which is a sensible ceiling until measurement improves.
Read more: OpenAI’s measurement gaps are keeping ChatGPT ads budgets at test level, Digiday, 1 October 2026
Also worth your time this week:
Web Development & Design: Google Finally Put Timescales On Technical SEO. Plan Around Them.
Search Engine Journal reports that Google’s Gary Illyes showed timing ranges for common Search processes at Search Central Live Deep Dive Europe in Barcelona. The recap gives “about 20 hours as the typical time for Google to discover a new URL and about 30 days to refresh one it already knows.” “Title and snippet updates take one to two days in the typical case,” while “Manual action removal takes one to two weeks in the typical case and four to six weeks at the slowest.” For anyone waiting on a recovery: “Typically, it takes about three to six months to recover,” with “a slowest case of six months to a year.” SEJ notes the figures come from an attendee’s recap and that no sample size or definition of “typical” was given.
For marketing leaders, these are the numbers that set realistic expectations with the board. A site migration isn’t a weekend job, and “A site move that remains unresolved at three months is at the high end of the typical timeframe.” Quality matters as much as code: “Five of the slowest times in the recap’s tables include ‘never.'” Google’s John Mueller made the same point on the Search Off the Record podcast, saying “the crawl demand is very often based on the perceived quality of a website” and “it’s not purely a technical thing.” He also played down llms.txt for AI crawlers: “I think the hope is bigger than the reality.” Strong technical foundations get you found. Quality content keeps Google coming back.
Why this matters:
- Google typically discovers a new URL in about 20 hours but takes about 30 days to refresh known pages, so plan content updates accordingly.
- Core update recovery typically takes three to six months, so set board expectations before promising a quick fix.
- Five of the slowest times include “never”, which shows quality issues can block indexing entirely.
Read more: Google Shows How Long Crawling, Indexing & Recovery Can Take, Search Engine Journal, 4 October 2026
Also worth your time this week:
Brand & Connection: Your Promotions Are Probably Losing Money. You Just Can’t See It.
The Media Leader reports on research by Les Binet and Kantar’s Dom Boyd, presented at the IPA Effectiveness Conference in London: “three-quarters (75%) of marketers are unable to estimate what proportion of their promotions are profitable.” Based on “a Censuswide survey of 250 senior marketing and insights leaders from UK consumer goods businesses,” it found “76% of marketers repeat promotions even when profitability has not been assessed,” and “four-fifths (81%) of marketers evaluate promotions over periods of less than 13 weeks.” Meanwhile, “a Nielsen analysis of 300m price promotions found that nearly two-thirds of such efforts actually lose money for businesses.” Binet didn’t hold back: “Price promotions are like class A drugs: expensive, dangerous and addictive.”
This is the brand version of the week’s measurement problem. Short windows make promotions look good and brand building look slow, so budgets drift towards the activity that’s easiest to count. Yet “the vast majority (91%) of marketers acknowledged it would be “easy” or “very easy” to shift funds away from price promotions and into long-term brand advertising if it were shown to be more profitable in the long term.” Separate research points to faster ways to prove brand is working. A Thinkbox study found ad awareness “is responsible for an average 17% of brand consideration,” and Lumen found “cumulative attention is 8x more impactful at delivering brand lift than mere frequency.” Binet’s advice is blunt: “identify which promotions lose money, cut them out, and reinvest the money in brand advertising.”
Why this matters:
- 76% repeat promotions without assessing profit, so audit last year’s promotions before planning the next one.
- 81% judge promotions over less than 13 weeks, which hides the long-term cost to pricing power.
- 91% say moving money into brand would be easy, so the barrier is evidence, not willingness.
Read more: Are you in a promotion dependency spiral? 75% of marketers are unable to identify profitable promotions, The Media Leader, 7 October 2026
Also worth your time this week:
AI & Automation: Agents Now Work For Days. Someone Still Has To Check The Work.
Digiday reports that at S4 Capital’s Monks, “Some runs go on for 36 to 48 hours and sizable projects can take several days,” according to chief AI and revenue officer Wesley ter Haar. “Monks is adapting the agentic setup into its workflow for social and performance marketing.” Agents “can turn out 50 ideas on brand and on brief in 10 minutes,” but when marketers ask for something bolder, “maybe only five are OK.” The economics are changing too. Tokens now sit in what ter Haar calls “the talent budget,” and teams keep asking for “more tokens, more tokens.” Some clients want every agent in one registry so they can “turn them off and on.”
The speed is real, but so is the need for judgement. Monks is even discussing whether to “reintroduce some friction” so people still add their point of view. Platforms are moving the same way. Social Media Today reports that Meta’s business assistant is gaining “agentic capacity, which will enable the business assistant to generate creative elements, update targeting and even change an ad’s budget using conversational prompts.” And regulation is catching up: Search Engine Journal reports OpenAI will add an invisible watermark to qualifying ChatGPT text in the EU, with EU AI Act rules giving existing systems “until December 2 to meet the marking and detection obligation.” AI can now do more of the work. Deciding what it’s allowed to change, and checking what it produces, is still a human job.
Why this matters:
- Agents running for 36 to 48 hours mean work moves faster than traditional sign-off processes.
- Only five in 50 bolder ideas were usable, so human creative judgement still decides quality.
- Meta’s assistant can now change targeting and budgets by prompt, so agree who approves those changes.
Read more: At S4 Capital’s Monks, agents can run autonomously, unsupervised for days, Digiday, 6 October 2026
Also worth your time this week:
Case Study of the Week: +54% ROAS, With 17% More Sales On 24% Less Spend
When platform numbers get harder to trust, efficiency you can prove matters more. Ellenvale Interiors is an online soft furnishings retailer selling upmarket cushions, throws, bedding and tableware in an English country style. After a strong Black Friday and November, budgets stayed at peak levels into quieter trading without any change to bidding, and ROAS slipped. Our brief was to recover ROAS and protect revenue without simply pulling budget and shrinking the account.
We moved every campaign to Target ROAS, raising the target by 10% each time a campaign hit it, and launched a dedicated brand campaign so brand and non-brand search were measured separately. We added long-tail luxury and interior design ad groups, plus campaigns for specific cushion colours and the outdoor range, and ran a feed-only Performance Max campaign on the top products by ROAS and revenue. Weak asset groups were paused and budget moved to the campaigns that were working.
- +54% improvement in ROAS
- +17% increase in conversions
- +16% increase in revenue
- 35% improvement in cost per conversion
In short, Ellenvale generated 17% more sales on 24% less spend.
Read the full Ellenvale Interiors case study.
Bringing It Together
Every story this week comes back to the same question: can you trust the number in front of you? Apple can now remove trackers from your data whenever it likes. ChatGPT reported five times the clicks an agency could find. Amazon says a fifth of your value lands after the window closes. Google wants proof that a human checked your AI content and that your experts actually exist. And three-quarters of marketers can’t say which of their promotions make money.
The answer isn’t to stop measuring. It’s to measure what you own and what you can prove. That means first-party and server-side tracking, platform numbers cross-checked against your own, longer evaluation windows and real people standing behind your content. The brands that do this will make better decisions while everyone else argues about whose dashboard is right.
Three Takeaways for the Week Ahead
- Fix your tracking before Apple breaks it further. Apple can add companies to its blocklist whenever it likes, across every iPhone browser. Audit your pixels and ID partners, move key tags server-side and grow the first-party data you collect directly.
- Put a human between AI and the publish button. Google now says to manually fact-check AI content, metadata included, and calls fake author profiles “a form of deception”. Name real experts on your content and check every AI-written title and description before it goes live.
- Stop judging spend on the shortest window you can find. 81% of marketers evaluate promotions over less than 13 weeks, and ChatGPT conversions can arrive late or not at all. Measure over longer periods, cross-check platform data against your own, and keep new channels at test budget until the numbers hold up.
Want help putting any of this into action?
We work with ambitious brands across SEO, PPC, web development, and content strategy. If this week’s news has raised questions about how reliable your tracking is, whether your paid media is being measured properly, or how to make sure your content earns Google’s trust, we’d like to talk.
Book a call with The Digital Maze.
Board Analysis is The Digital Maze’s weekly marketing briefing, published every Friday. It covers search, paid media, web development, brand, and AI for marketing managers, directors, and leaders who need to stay ahead without wading through every trade publication themselves.
