Anthropic now watermarks AI text under EU law, ChatGPT ads miss the mark 14% of the time, and three SEO research pieces show what actually drives AI visibility.
Hi Marketers,
Welcome to the new issue of HighDegree*: Cutting Through the Noise in SEO & Digital Marketing
Two big platform moves this week have direct consequences for how you create, publish, and rank content. Anthropic has started watermarking AI-generated text under EU law. ChatGPT is running ads on a quarter of commercial queries, and early data shows the targeting has real problems. Three separate research pieces also make the same argument from different angles: the SEO tactics you already use are your best bet for surviving whatever Google does next.
Let’s jump in.
➞ Digital Marketing This Week
- Anthropic Now Watermarks AI-Generated Text and Files Under EU Law
- ChatGPT Runs Ads on 26% of Commercial Queries, But 1 in 7 Miss the Topic
- How to De-Risk Your SEO Strategy When Google Keeps Changing the Rules
- What Google’s Documentation Rewrite Teaches You About Better Content
- These SEO Strategies Drive 90% of Your AI Visibility
➞ Subscribe
If you are not subscribed to HighDegree* yet, please subscribe to us by clicking the “Subscribe” button below. I will post every issue to HighDegree* Substack version, so you can get it instantly in your inbox every time I hit publish!
➞ Anthropic Now Watermarks AI-Generated Text and Files Under EU Law
The EU’s AI Act Transparency Code took effect on August 2, 2026, and Anthropic is already moving on it. Every Claude model launched on or after that date now embeds an invisible watermark directly into generated text. You won’t see it in the output, and it doesn’t change what Claude writes, but it travels when you copy and paste, and may survive light editing.
For files like PNGs, JPGs, and SVGs, Claude attaches signed provenance metadata using the C2PA open standard, the same format adopted by Google, Meta, Microsoft, and OpenAI. The watermark applies everywhere Claude runs: the API, Claude.ai, Claude Code, Claude Cowork, and Claude Tag. Older models are getting the same treatment in a later rollout.
If someone runs a detection check on content you’ve published, a Claude mark signals it was processed by Claude. That doesn’t mean Claude wrote the whole thing; the mark appears even if you used it to proofread or translate a draft. But that distinction may not be obvious to readers or employers. Review platforms and publishers already have AI disclosure tools in place. This watermark just gave them a machine-readable trigger to use them.
Read the full report from support.claude.com ➞
The ad picture inside ChatGPT is getting complicated fast, and the numbers below explain why marketers need to pay close attention before they spend a dollar on it.
➞ ChatGPT Runs Ads on 26% of Commercial Queries, But 1 in 7 Miss the Topic
SE Ranking analyzed 50,006 commercial prompts across 20 niches in the US and found that ChatGPT served an ad on 25.94% of them, close to the 29.45% rate in Google’s AI Mode. That’s a meaningful number for a platform still in early rollout.
The ad format is currently a single sponsored card below the generated answer, with no competing ad beside it. OpenAI is already testing multi-advertiser placements, so that advantage may not last.
Relevance is the bigger problem. About 14.35% of ads had no meaningful connection to the query they ran against. In the Relationships niche, 51.1% were off-topic. In News and Politics, the mismatch hit 54.2%. The system uses natural-language context hints instead of keyword targeting, which gives it flexibility but also lets it stretch too far.
For B2B advertisers there’s a structural issue on top: ChatGPT ads only reach Free and Go plan users. Customers on Plus, Pro, Business, and Enterprise see no ads, cutting out a large portion of professional buyers.
And in 96.37% of ad placements, the advertiser was not cited as a source in the answer directly above the ad. Paid placements and organic AI citations are separate challenges that require separate approaches.
Read the full report from seranking.com ➞
Knowing how the ad systems work in AI platforms is one piece. The bigger question is whether your SEO strategy can hold up when Google keeps changing what it rewards.
➞ How to De-Risk Your SEO Strategy When Google Keeps Changing the Rules
Dan Taylor at Search Engine Land wrote a detailed guide on building an SEO strategy that holds up when Google keeps shifting what it rewards. Most ranking volatility isn’t random. It comes from specific, identifiable factors you can actually address.
Modern Google runs on machine learning and neural networks, not fixed rules. Rankings shift in response to real user behavior signals like Chrome data and NavBoost clickstream data. Quick drops are more often a UX problem than an algorithm problem.
The guide’s practical advice breaks into four areas. Keep your technical setup clean so Google can crawl and index without friction. Build content around E-E-A-T (Experience, Expertise, Authoritativeness, and Trustworthiness), avoid publishing raw AI-generated content at scale, and cover topics from multiple angles to hedge against intent shifts. Improve UX to cut bounce rates and pogo-sticking, which Navboost tracks over a 13-month window. And earn links through digital PR and original research instead of buying them.
The guide draws a clear line between low-risk tactics that hold through updates and high-risk tactics that invite penalties. For bloggers and small business owners, the takeaway is that building real authority compounds over time in a way that chasing loopholes never does.
Read the full report from searchengineland.com ➞
If you want a concrete example of how editorial quality signals translate to better SEO, Google’s own documentation just became a case study worth studying.
➞ What Google’s Documentation Rewrite Teaches You About Better Content
Roger Montti at Search Engine Journal found a practical content lesson buried in Google’s recent crawl budget documentation update. Google rewrote the guide to be more precise and useful, and the editorial moves they made are worth copying on your own site.
Montti identifies three improvement types. First, Google swapped general descriptions for specific ones. “Responds quickly for a while” became “responds consistently and its response times remain stable or improve.” “Server errors” became “5xx HTTP status codes or HTTP 429.” Second, they cut explanations of why Google does something and replaced them with what actually happens, removing filler and making sentences do more work. Third, they eliminated what Montti calls “reading road bumps,” sentences that introduce two clashing ideas and force a mental stop.
Google’s natural language systems were built to approximate human comprehension, so content that reads cleanly to people gets processed more clearly by AI too. Specific language, direct sentences, and clean structure are ranking signals. A targeted audit applying these three types of precision to existing posts is often more effective than writing new content from scratch.
Read the full report from searchenginejournal.com ➞
The same principle runs through our final article: the SEO fundamentals you already know are doing more work in AI search than most people realize.
➞ These SEO Strategies Drive 90% of Your AI Visibility
Cyrus Shepard of Zyppy makes the case that most businesses can hit 90% or more of their AI visibility goals through well-executed traditional SEO, without a separate Generative Engine Optimization (GEO) strategy.
Shepard breaks AI visibility into three goals: getting content eligible for AI crawlers, earning AI citations, and getting recommended in AI answers. The most effective levers for each are already part of standard SEO.
On crawlability: audit your robots.txt for AI bot access, check CDN and firewall settings, and confirm content is visible without JavaScript rendering. Many sites block AI crawlers through broad bot rules set by technical teams who never looped in marketing.
On citations: rank for query fan-outs, the background sub-queries AI engines run when forming answers. Format headings around specific questions and back claims with data to become a stronger citation candidate.
On recommendations: keep brand facts consistent across every surface you control, build profiles on third-party platforms, encourage detailed customer reviews, and invest in digital PR focused on authentic mentions rather than keyword-rich links.
For most site owners, strong technical hygiene, topical authority, and earned mentions cover most of the ground. Advanced GEO tactics are worth exploring once you’ve built that base.
Read the full report from signal.zyppy.com ➞
➞ From Google
Everything from Google search this week —
→ Google’s Gemini App Surges to 1 Billion Users (techcrunch.com)
→ Google Has Silently Canceled Gemini 3.5 Pro, Says Semi Analysis Report (officechai.com)
→ Google Confirms Testing New Buttons on Home Page to Drive Users to AI-Powered Search Features (searchengineland.com)
→ Google Search Console Generative AI Performance Report Expands to More (seroundtable.com)
→ Google Brings New AI Agent Capabilities to Ads and Analytics (searchengineland.com)
→ Google Tests Advertiser List Above Mobile Search Ads (searchengineland.com)
→ Google Ads Sponsored Results Tests Advertiser Name and Favicon Summary (seroundtable.com)
→ Google Travel Campaigns Migrating to Search Campaigns for Travel (seroundtable.com)
→ Google Adds Campaign Data Import Validation Report (searchengineland.com)
→ Google Still Aggressively Poaching Agency Ads Accounts (seroundtable.com)
→ Google AI Labels on Local Pack and Google Discover Ads (seroundtable.com)
→ Google Local Disallows Repeated Bilingual Names and Transliterations (seroundtable.com)
→ Google: Hreflang URLs Not Indexed in the Proper Sense (seroundtable.com)
→ Google’s Ex-AI Chief Jeff Dean Explains How to Improve Context Engineering (searchenginejournal.com)
→ Google Search Ranking Volatility Heats Up Into August 5th and 6th (seroundtable.com)
→ Google Maps Adds Agentic Features, Including Food Ordering and Hotel Bookings (techcrunch.com)
→ Google Crawlers Send HTTP Requests: HEAD, OPTIONS, PUT, PATCH, DELETE (seroundtable.com)
→ Google to Auto-Upgrade Automatically Created Assets to AI Max (searchengineland.com)
→ Google Ads Won’t Let Free Email Accounts Take Sensitive Actions (seroundtable.com)
→ New Google Product Listing Ads Deal Ends Experimental Annotation (seroundtable.com)
→ Google Ads Tests Anchor Text Hyperlinks in Ad Descriptions (seroundtable.com) → New Google Ads Report on New Customers Acquired (seroundtable.com)
→ Google Ads Live Tests of Strong Match Labels (seroundtable.com)
→ How Gemini Plans Such Detailed Vacation Itineraries for You (blog.google)
➞ AI + Social
Find out what’s happening in the social media and artificial intelligence world —
→ Tech Industry Is Buzzing After a Claude Agent Hacked into a Gym (techcrunch.com) – A Claude-based AI agent found and exploited a real vulnerability in a gym’s booking system, reigniting debate about AI agent security and autonomous systems running in production.
→ Anthropic Is Turning Claude Code’s Auto Mode on by Default (techcrunch.com) – Anthropic is enabling auto mode by default in Claude Code, letting the coding agent take actions without waiting for user confirmation at each step.
→ ChatGPT Can Now Make You a Dinner Reservation (theverge.com) – OpenAI has integrated restaurant booking and waitlists into ChatGPT through a partnership with Yelp, making agentic real-world tasks available directly in chat.
→ ChatGPT’s Search Index Serves Small Sites Too, Data Shows (searchenginejournal.com) – New data suggests ChatGPT’s search index does not exclusively favor large publishers, which opens realistic citation opportunities for smaller sites and niche blogs.
→ OpenAI Launches ChatGPT Desktop App for Linux (techcrunch.com) – OpenAI has released a native ChatGPT desktop app for Linux, expanding platform reach to developers and technical users who rely on Linux environments.
→ As AI-Led Attacks Multiply, OpenAI Launches a New Cyber Model (techcrunch.com) – OpenAI has released a specialized AI model for cybersecurity, designed to help security teams detect and respond to the growing volume of AI-generated attacks.
→ OpenAI Acquires Presentation Startup NextSlide (techcrunch.com) – OpenAI acquired NextSlide, a startup focused on AI-powered slide creation, signaling its push into productivity and document creation tools.
→ Spotify Will Label AI Persona Profiles and Exclude Their Music from Recommendations (techcrunch.com) – Spotify is adding disclosure labels to AI-generated artist profiles and will stop surfacing their music in algorithmic recommendations.
→ Half of US Shoppers Verify AI Recommendations on Reddit Before They Buy (business.reddit.com) – Reddit research found that 50% of US shoppers cross-check AI-generated product recommendations on Reddit before making a purchase, underlining Reddit’s role in the buyer decision process.
→ X Updates Creator Revenue Share Parameters (socialmediatoday.com) – X has changed how it calculates creator revenue, affecting payouts for many accounts and drawing concern from creators who rely on ad share income.
→ YouTube Announces Changes to Creator Monetization (socialmediatoday.com) – YouTube is updating monetization requirements, raising the watch hour threshold new creators need to reach before they can start earning from their channels.
→ Mark Zuckerberg’s AI Manifesto Is Exactly Why People Don’t Like AI (techcrunch.com) – Zuckerberg published a vision for AI social companions that critics say sidesteps real concerns about privacy, trust, and the role of AI in personal relationships.
→ Meta’s Lack of Consumer Trust Could Derail Its AI Plans (socialmediatoday.com) – Analysts warn that Meta’s persistent reputation gap on user trust may slow adoption of its AI-powered features, even as the company accelerates its AI product rollout.
→ Meta Has a New Open-Weight AI Model (theverge.com) – Meta released a new open-weight language model, continuing its strategy of publishing model weights publicly to compete with closed offerings from OpenAI and Anthropic.
→ Meta Adds Dedicated Ad Exclusion Audiences (socialmediatoday.com) – Meta is giving advertisers more control over who does not see their ads, adding structured exclusion audiences to its ad targeting interface.
→ Meta/Facebook Crawling Web and May Be Building Search Engine (seroundtable.com) – Reports suggest Meta is significantly increasing its web crawl activity, prompting speculation that the company is developing a standalone search engine.
→ OpenAI and Four Rivals Just Agreed on One Standard for AI Agents (thenextweb.com) – OpenAI and several competitors have aligned on a shared open standard for AI agent interoperability, a move that could simplify how AI agents communicate across different platforms.
→ ChatGPT Brings Unlimited Text Chats to Free Users (techcrunch.com) – OpenAI has removed the daily text chat limit for free ChatGPT users, making the service more accessible ahead of further ad-supported tier expansions.
→ LinkedIn Updates Feed Display to Drive More Post Replies and Comments (socialmediatoday.com) – LinkedIn is testing a new feed layout that emphasizes replies and comment threads, pushing users to engage more deeply with posts rather than scrolling past them.
→ YouTube Adds Vertical Livestream Practice Mode (socialmediatoday.com) – YouTube introduced a practice mode for vertical livestreams, letting creators test their setup and delivery before going live to their audience.
→ Cloudflare Launches Kitesurf, a Browser Built for AI Agents (techcrunch.com) – Cloudflare released Kitesurf, a headless browser designed specifically for AI agents to navigate the web, handle authentication, and complete multi-step tasks at scale.
→ Improving GPT-5.6 Sol in ChatGPT and Expanding Access to GPT-5.6 Luna for Free Users (openai.com) – OpenAI detailed updates to GPT-5.6 Sol for paid users and announced that GPT-5.6 Luna, a lighter model, is now available to free-tier ChatGPT accounts.
➞ Worth Reading
These are the articles that will help you refine your marketing knowledge —
→ See Branded and Non-Branded AI Queries in the Clarity Citations Dashboard (clarity.microsoft.com)
→ Does Google Use User Behaviour, Website Design and Layouts in SEO Rank Assessments? (assertive-media.co.uk)
→ The Future of Discoverability (leadershipinseo.com)
→ How I Use Knowledge Graphs on Content Sites (fajela.com)
→ Content Prioritization in an AI Search Era: A Framework and Worksheet (aleydasolis.com)
→ How to Communicate and Sell AI Search Optimization to Decision Makers (youtube.com)
→ How cats.txt Showed llms.txt Evidence Is GEO Astrology (markwilliamscook.substack.com)
→ What 50k Query Fan-Outs Reveal About Brands (moz.com)
→ 10 SEO Use Cases for Auditing Your Accessibility Tree for AI Search (searchengineland.com)
➞ What This All Means for Your Strategy
Anthropic’s watermarking move means AI-assisted content is now potentially machine-tagged everywhere it lands. That’s not a reason to stop using AI tools, but it is a reason to be deliberate about where that content shows up under your name and how much you edit it before publishing.
The ChatGPT ads data is a caution for advertisers. With 14% of placements off-topic and professional buyers mostly on ad-free plans, jumping in without careful audience verification is an easy way to burn budget. Organic AI visibility and paid placements need separate strategies.
The three SEO pieces all point at the same thing from different angles: precision, clarity, topical authority, and technical cleanliness are what hold up through Google’s changes. Montti’s editorial analysis, Taylor’s de-risking guide, and Shepard’s AI visibility research all arrive at the same conclusion. Writing that is specific and well-structured is what ranks and gets cited. That’s the part of your strategy worth compounding.
Until next week,
— Nishat from HighDegree* Marketing.
P.S. Have a question about implementing these strategies? Hit reply – we read every email and often feature reader questions in future issues.
➞ Who is Nishat Shahriyar?
I’ve been a Digital Marketing Strategist since 2007. Today I’m the Product Marketing Lead at Fluent Forms (a WordPress form and lead-generation plugin used by 700K people), FluentPlayer (a WordPress Video Player Plugin) and Ninja Tables (a WordPress table builder plugin used by 80K people). Before that, I led marketing at Fluent Support (a WordPress helpdesk plugin used by 10K people).
Connect with me on LinkedIn or follow me on X/Twitter at @rednishat for SEO & Marketing updates.