There is a version of this week where AI is the whole story. Google is testing AI-written summaries under its ads. OpenAI is building out video and conversational ad formats inside ChatGPT. Major publishers are rebuilding their websites so machines can read them faster than people can. On paper, it looks like the industry's AI transition just accelerated another notch.
Then the actual performance data landed, and it told a different story. The best-performing brand campaigns of H1 2026 were not the most AI-polished. They were the most honest. Burger King ditched its mascot and its usual gloss for a stripped-back, consumer-first relaunch and posted 5.8% comparable sales growth in Q1 2026. Dove let real, unfiltered Reddit reviews of its products stand in as the campaign itself and picked up six Cannes Lions for it. Clue, the women's health app, built its whole marketing strategy around ordinary users telling their own stories rather than creators or celebrities, and generated 30,000 sign-ups in a single day for one values-led campaign.
That is the tension running through everything this week. AI is quietly rewriting how brands get found, how ads get written, and how websites get built. But the brands actually winning with customers right now are the ones using AI to buy back time for the human work, not to replace it.
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.
Search & Intelligence: Google Decided Who Gets Credit In AI Search
Search Engine Land published a distinction this week that every marketing team optimising for AI visibility needs to understand: there is a difference between AI engines "using" your brand information and formally "citing" it. The research found that ChatGPT retrieves a near-even split of cited URLs (16.57 on average) and uncited URLs (16.58) per response. In other words, roughly half the sources an AI model pulls from never get named or linked. Your content can be feeding an AI answer word for word and your brand can still be invisible to the person reading it.
The pattern behind who gets cited and who gets quietly absorbed is not random. AI engines are far less likely to cite content that simply restates what everyone else in a category already says. Generic, safe, consensus content gets used as background knowledge. Original analysis, direct data, and content that says something no one else has said gets the citation and the click. For marketing teams treating "AI SEO" as a rewrite of existing content into more AI-friendly formatting, this is the uncomfortable correction: formatting will not fix a content strategy that has nothing new to say.
Google made a smaller but practically useful move in the same direction this week, adding a new "platform properties" option to Search Console that lets brands and creators track how their Instagram, TikTok, X, and YouTube content performs directly within Google Search and Discover, even without owning a website. It closes a measurement gap that has made off-site social and video performance largely invisible inside Google's own reporting tools, and it works via verification even for creators with no website of their own.
Why this matters:
- ChatGPT pulls almost equal numbers of cited and uncited sources per response, so being "used" by AI is not the same as being credited. Track both, not just one.
- AI systems rarely cite generic content that just repeats category consensus. If your content strategy is built on covering the same ground as competitors, it is structurally unlikely to earn citations.
- The new Search Console platform properties give a free, consolidated view of how your social and video content performs in Google Search, closing a reporting gap many teams have patched with manual spreadsheets.
Read more: Used or cited: The two ways brands appear in AI search, Search Engine Land, 7 July 2026
Also worth your time this week:
Paid Media & Performance: Google Wants To Write Your Ad Copy Too
Google is testing a feature that hands over a slice of message control that advertisers have always owned outright: AI-generated summaries appearing directly beneath ad descriptions in Search results. The test, first spotted by marketer Darcy Burk, shows Google's own AI writing contextual copy under a paid ad, separate from whatever the advertiser wrote. The summaries carry a visible disclaimer that the AI "can make mistakes," which is a fair warning but a strange one to attach to something sitting directly beneath a brand's paid message.
For marketing directors who treat ad copy as a carefully controlled brand asset, this is worth watching closely rather than dismissing as a minor UI tweak. If Google's AI is willing to add unsolicited context to a search ad, brand safety and message accuracy reviews may need to extend beyond the copy a team actually writes and submits.
Two other Search Engine Land pieces this week were less about new features and more about diagnosing where paid budgets are actually being wasted. The first argued that account structure, not bid strategy, is the real lever behind Google Ads performance in 2026, since Smart Bidding needs a reliable 30 to 50 conversions a month per campaign to exit its learning phase properly. Over-segmented accounts cause bids to cannibalise each other across overlapping auctions, and the practical fix is consolidating underperforming campaigns before touching bids at all. The second broke down how competitors are targeting branded search traffic without ever naming a brand directly in their ad copy, using dynamic keyword insertion, comparison landing pages, and brand modifier terms like "[Brand] alternative" to quietly divert clicks and inflate CPCs on defensive brand bidding.
Why this matters:
- Google's AI, not your team, may soon write the context shown beneath your ad copy, carrying its own disclaimer about accuracy. Brand safety reviews may need to extend beyond copy your team actually submits.
- Smart Bidding needs 30 to 50 conversions a month per campaign to work properly. Over-segmented accounts are quietly wasting budget through bid cannibalisation long before the algorithm gets blamed.
- Competitors can target your branded traffic through comparison pages and brand modifier keywords without naming you once, which means your defensive bidding spend may need a fresh audit.
Read more: Google tests AI-generated summaries in Search ads, Search Engine Land, 1 July 2026
Also worth your time this week:
Web Development & Design: Publishers Are Rebuilding The Web For Robots
Digiday reported this week on a genuinely structural shift in how major publishers are architecting their websites. Time and The Economist are converting parts of their sites into markdown specifically so AI agents and crawlers can process them faster and more cheaply than scraping raw HTML. The numbers are stark: TollBit, the infrastructure provider behind Time's conversion, reports a 90% reduction in the tokens needed to process a page in markdown versus HTML, and a drop in processing time from over a minute scraping raw HTML down to 0.25 seconds. The Economist, meanwhile, is piloting WebMCP, a new agent-accessible content standard backed by Google and Microsoft, for its non-paywalled marketing materials.
This matters well beyond publishing. It is an early, concrete signal that technical content architecture, not just copy quality, is becoming a genuine competitive layer for AI and GEO visibility. If AI agents increasingly prefer to fetch a lightweight markdown version of a page over scraping bloated HTML, sites that make that easy will get processed, cited, and represented more accurately than sites that do not. For any marketing team weighing up whether "AI readiness" is a content problem or a development problem, this week's evidence says it is both.
On a smaller but more immediately practical note, Google fixed a Search Console indexing report that had been stuck showing stale data from 11 June for three weeks, leaving developers and SEO teams unable to properly diagnose indexing issues on their own sites during that window. The report now reflects data through to 29 June.
Why this matters:
- Time's markdown conversion cuts AI processing time to 0.25 seconds a page, down from over a minute scraping raw HTML, according to TollBit's own reporting.
- A 90% reduction in tokens needed for AI agents to process markdown versus HTML suggests technical architecture is becoming a real lever for AI visibility, not just content quality.
- The Economist's WebMCP pilot signals a genuine emerging standard for agent-readable content, not just a one-off workaround, worth tracking as it matures.
Read more: How Time and others are rebuilding parts of the web for AI agents, Digiday, 1 July 2026
Also worth your time this week:
Brand & Connection: Radical Honesty Is Outperforming AI Polish
Marketing Week's interview with Clue CMO Louise Troen was the standout brand story of the week. The women's health app has built what it internally calls a "radical human marketing" strategy, built around real users telling their own stories rather than creators or celebrities. "When I joined Clue, the first thing I did was talk to our community," Troen told Marketing Week. "I asked them what had benefited them from the product, how it had changed their life and how we could do better as a company and a brand."
The commercial case for this approach is not just a feeling. When Clue offered free premium subscriptions to women in US states where abortion access had become more restricted, the campaign generated around 30,000 sign-ups in a single day. Troen deliberately refuses to attach an ROI figure to trust or credibility as a metric: "If you try and measure trust or credibility, all of a sudden you're moving into the performance space and away from purpose. My responsibility is to make sure we don't put a number on integrity." Even Clue's influencer approach reflects the same principle, favouring grassroots user-generated content creators over paid influencers, on the basis that "it was our grassroots UGC creators who performed best," and putting those advocates directly on the payroll rather than treating them as free content sources.
Marketing Dive's review of the best brand campaigns of H1 2026 reached the same conclusion from a completely different angle. Burger King's mascot-ditching, stripped-back relaunch drove 5.8% comparable sales growth in Q1 2026, with 97% of franchisees backing higher ad fund contributions off the back of it. Dove let unfiltered Reddit reviews of its products run as the campaign itself, warts and all, and won six Cannes Lions for the approach. Neither campaign is more "creative" in a traditional production sense than a heavily art-directed, AI-assisted alternative would have been. Both won by removing the layer of polish that usually sits between a brand and its customers.
Why this matters:
- Clue's free-subscription campaign for women in restricted-access states generated around 30,000 sign-ups in a single day, evidence that values-led honesty converts, not just builds sentiment.
- Burger King's honesty-led relaunch drove 5.8% comparable sales growth in Q1 2026, direct evidence that trust-building creative can move the P&L, not just brand tracking scores.
- Dove's unfiltered Reddit reviews campaign won six Cannes Lions, showing that loosening creative control can outperform tightly scripted, AI-polished messaging.
Read more: Why women's health app Clue is betting on "radical human marketing", Marketing Week, 1 July 2026
Also worth your time this week:
AI & Automation: Ranking Number One Means Nothing To The AI Overview
Marketing Dive published research this week that should reset how marketing teams think about the relationship between traditional SEO and AI visibility. The data found that 42% of brand citations appearing in standard organic search results simply do not carry over into AI-generated overviews at all. Just as strikingly, 28% of brands that do get cited in AI overviews are missing from the organic results for the same query entirely. These are not overlapping systems measuring the same thing from two angles. They are two separate visibility contests, and a strong showing in one carries no guarantee of a showing in the other.
The research also found that content type matters more than most SEO playbooks currently account for: YouTube content is 4.3 times more likely to appear in AI overviews than equivalent organic search results. Category-level queries, the broad "best X for Y" searches rather than specific branded ones, showed the highest visibility gap, with brands appearing in 63.9% of AI overviews for those queries compared to 55.7% in organic results. For any team that has built its SEO strategy purely around organic ranking factors, this is direct evidence that AI overview visibility needs its own distinct measurement and its own distinct content approach, particularly around video and broad category coverage.
OpenAI is also moving to make ChatGPT a genuine paid advertising surface rather than a simple headline-and-link unit. Digiday reported the company is actively hiring engineers to build image, video, native, conversational, and interactive ad formats inside ChatGPT, with several of the roles explicitly built around privacy and safety compliance. For marketing teams currently treating ChatGPT purely as an organic AI visibility channel, this is a signal that a genuine paid channel is coming, likely with its own creative specs and its own trust constraints to navigate from day one.
Why this matters:
- 42% of brand citations in organic search results are missing entirely from AI overviews, meaning strong SEO rankings offer no guarantee of AI visibility.
- YouTube content is 4.3 times more likely to appear in AI overviews than organic search results, a strong argument for treating video as core AI visibility infrastructure, not a side channel.
- OpenAI is actively building richer ad formats inside ChatGPT, with roles built around privacy and safety compliance, signalling a genuine paid AI channel is approaching fast.
Read more: How brands can improve chances of showing up in AI search overviews, Marketing Dive, 1 July 2026
Also worth your time this week:
Case Study of the Week: 1,330% Growth In Search Visibility
This week's theme, that the fundamentals still win even as AI reshapes the surface, is exactly what played out for Wetroom Materials, an award-winning wetroom solutions provider serving architects and house builders that came to The Digital Maze for SEO support.
We overhauled on-page SEO across meta titles and descriptions, sharpened blog and service page content to align more closely with real search intent, and built out a stronger internal linking structure across the site. We also ran an experimental A/B test on emoji use in meta titles specifically to test whether small, testable changes could shift clickthrough rate, the same kind of disciplined, evidence-led approach that this week's research on AI citations rewards.
- +1,330% increase in search visibility over 20 months
- +12% year-on-year increase in URLs featuring ranking keywords
- +19% increase in site clicks within six months
- +18% increase in organic blog traffic through content optimisation
- +25% increase in clickthrough rate from emoji testing in meta titles
Read the full Wetroom Materials case study.
Bringing It Together
The thread running through this week's news is not "AI versus humanity." It is a much more useful, much more actionable distinction: AI as infrastructure versus AI as substitute. Used as infrastructure, AI is genuinely reshaping how content gets discovered, how ads get built, and how websites get architected, and teams that ignore those shifts will fall behind on distribution even if their creative and their offer are strong.
Used as a substitute for the human judgement and honesty that actually build trust, AI is quietly costing brands the thing that is converting best right now. Burger King, Dove, and Clue did not win this week by being the most technologically advanced. They won by being the most honest, and by trusting real customers and real employees to carry the message rather than a heavily produced, algorithmically optimised version of it.
The practical takeaway is not to choose one over the other. It is to be deliberate about which is which. Use AI to fix the plumbing: your content architecture, your measurement, your ad production efficiency. Do not let it write the parts of your brand that are supposed to sound like a person.
Three Takeaways for the Week Ahead
- Audit whether your brand is "used" or "cited" in AI search results. Run a handful of category and branded queries through ChatGPT, Perplexity, and Google AI Mode this week. If your data or language is clearly informing the answer but you are not named or linked, your content needs to say something more original to earn proper credit, not just be reformatted for AI readability.
- Check your branded search terms in Google Ads for competitor activity. Search your own brand name and close variants and look for comparison pages or "[Brand] alternative" ad copy from competitors. If it is happening, your defensive branded bidding budget may need a fresh look, since the traffic diversion tactics documented this week rarely name a competing brand directly.
- Before greenlighting the next AI-polished campaign, ask whether it would survive being posted unfiltered on Reddit. Clue, Dove, and Burger King all won this week by removing a layer of polish, not adding one. If the honest, unedited version of your message would not hold up to public scrutiny, it is not ready to go out, however well the AI-generated version performs in testing.
Want help putting any of this into action?
Whether it is AI search visibility, PPC performance, brand positioning for tougher trading conditions, or making sense of what these platform changes mean for your specific strategy, we can help.
Get in touch with The Digital Maze. 
