Which of your topics are blurring together

Wave 304; How AI Search Chooses Passages; Cats.txt; Get Recommended; Local SEO AI Workflows; Topical Gravity; Claude's Watermarks; and Much More!

Author

Yoyao
August 15, 2026 • Est. Learning Time: 10 minutes

In partnership with

FIRST …

NFL preseason started. Which means the New York Jets rebuilt again in the offseason.

New players. New coaches. New optimism.

I have hope every offseason. But they still lose.

I've watched people run the same offseason on their own sites.

New content push. New AI workflow. New tool stack.

The surface gets swapped out. The outcome is no different from last year.

Every coach has a plan for how the team will play.

The box score won't tell him if it happened.

Three catches for 20 yards is a result, not a reason. But the film is where he finds out the receiver ran the wrong route.

Search Console is your box score.

It tells you which pages dropped and which queries slipped. But it won't show you the structure underneath.

So I built a free Content Audit tool that does.

Enter a domain. It reads up to 100 pages and groups them by topical meaning, then maps which topics are strong, which ones are blurring into each other and which pages don't belong anywhere.

The map lands in your inbox in about 10 minutes.

Run it on your site or on a competitor's.

Before you rebuild again this quarter, find out what you're actually working with.

P.S. Run it on a prospect's site before your next pitch call. You show up with the film. They've only ever seen the box score.

AD

Your voice. Every platform. No writing required.

You ghost your own socials by Wednesday. SureThing learns your voice and ships native posts to LinkedIn, X, Instagram, and TikTok, without you writing a thing.

SEO + GEO

Metehan Yesilyurt explains why AI-search visibility has three separate gates: retrieval, passage reranking, and final citation use. A strong URL can clear the first gate while its most relevant section loses to a more direct answer on another site.

  • Diagnose the failed stage first: No retrieval points to indexing, coverage, authority, or freshness. A retrieved passage that loses points to answer shape or local relevance. A strong passage without a citation can reflect source quality, diversity, or generation behavior.

  • Match the answer shape to the query: Use a shortlist with criteria for “best” queries, a like-for-like comparison for versus queries, a concise definition for entity questions, and ordered steps for how-to queries.

  • Put the answer before the supporting detail: The article’s tests found that strict rerankers gave a direct shortlist far higher scores than a product description. Add a clear answer-bearing sentence in the first two lines of the relevant section, then add evidence and nuance.

  • Test versions, not a made-up GEO score: Compare the same passage and competitors across a few model families. Track retrieval exposure, candidate strength, and citation or mention rate separately.

  • Write for local clarity, not a fixed chunk size: Systems divide content differently. Keep the subject, answer, and evidence close together so a section still makes sense when read on its own.

Mark Williams-Cook created a fake cats.txt standard, then showed it could clear the same four tests often used to sell llms.txt: AI bots crawled it, Google indexed it, AI answers repeated facts from it, and ChatGPT recommended it. Each result can happen because the file is an ordinary public URL, not because a system treats it as a special AI-search signal. His point is not that llms.txt can never matter. It is that GEO claims need a test that distinguishes correlation from an actual mechanism.

Shadrack Wanjohi ran the same B2B software buying prompt 11 times and found that appearing in an AI answer does not mean the model preferred the brand. The exercise separates presence in the shortlist from the recommendation that shapes a buyer’s next move.

  • Gong appeared in all 11 answers, but was the clear recommendation in only three.

  • The unchanged prompt produced four different readings of the buyer’s job: prospecting pointed to Apollo, coaching new SDRs to Gong, stack consolidation to Salesloft and Clari, and CRM migration to HubSpot and Outreach.

  • Zapier appeared in 70% of the participant-level source records. Oliv, Tellius, and ZoomInfo’s Pipeline each appeared in 60%.

  • A separate index of 144 answers across six models found one brand with 86% share of voice but none of the 36 first-pick recommendations.

  • Track mention rate and recommendation rate by buyer job, then compare the recurring third-party sources behind wins and losses.

Amanda Jordan separates local SEO tasks into automation, AI, and strategist work after managing workflows for more than 20,000 restaurant locations. Her approach filters inputs before the model sees them, assigns each agent one job, and holds publishing behind validation and human review.

My Take: It’s always interesting to see others’ workflows. The lessons are not the prompts. It’s the decision about where business context belongs, what data is safe to pass through the system, and which outputs require a strategist’s judgment before a client sees them.

Margarita Loktionova says AI search visibility is won across related questions, not one tracked prompt at a time. Semrush’s framework calls on teams to define the topics that matter, build coverage on and off site, measure their share against competitors, and expand into adjacent topics from a position of strength.

My Take: Floyi’s framework starts from a similar premise as topical gravity: a topical map is not a keyword list. It connects the audience’s questions, the brand’s positioning, and the commercial pages needed to earn trust. The work is then to build enough depth and internal connection that both people and machines can see the relationship.

Yulia Deda analyzed 50,006 commercial ChatGPT prompts across 20 US niches and found an ad market that can generate reach without reliably delivering relevance or organic visibility.

  • 12,974 of the 50,006 prompts returned an ad, a 25.94% ad rate.

  • Every observed placement was a single sponsored offer below the answer, with no competing ad beside it.

  • 14.35% of placements were no more related to the prompt than a random prompt-ad pairing. The mismatch rate exceeded 50% in Relationships and News & Politics.

  • Advertiser domains appeared among the answer’s citations in 3.63% of placements. Exact advertised URLs appeared in 0.09%.

  • SE Ranking’s own test generated more than 97,000 impressions and 1,263 clicks over roughly two weeks, but few sign-ups.

Barry Schwartz reports that AI-generated illustrations are starting to appear in Google AI Overviews. Inspired Taste saw step-by-step cartoons beside a spaghetti squash recipe that already included safety guidance, cooking details, and a video. The publisher’s concern is clear: a cited source can look as if it created an illustration it did not.

Matt G. Southern reports that Google amended its complaint against SerpApi after a federal judge dismissed its original DMCA claims over Search-result scraping in July. The new filing adds terms from Google's licensing deals to support its claim that copyright owners authorized SearchGuard. Claims involving results without copyrighted content cannot be revived, while discovery remains frozen until the court rules on another motion to dismiss.

Matt G. Southern reports that Resoneo reviewed 1,249 ChatGPT answers from July and found that unlicensed publishers were served through OpenAI's in-house index in the same format as licensed partners. In a separate paid thinking-mode sample of 16,407 results, Google scraping supplied about 75% while the in-house index supplied about 24%. Of 534 cited pages reviewed, 83.6% of the 463 pages with an H1 had that heading in the stored snippet.

My Take: There are still people out there saying ChatGPT has no index, not sure why since they’ve had it for some time now. Some are also saying that in order for a page to be cited in ChatGPT, that page must be indexed in Google or Bing. Not true again. I’ve done my own tests and mentioned on the IMG podcast that it’s not true. But, your content still needs to be good or it won’t get used.

Dan Taylor explains how fraudulent DMCA complaints can remove a page from Google before a person checks the claim. He recounts a case in which Search Engine Land and Press Gazette articles disappeared after a complaint cited images that Search Engine Land had not used. His response plan starts with the notice and Lumen Database record, then moves to CMS logs, Wayback snapshots, RSS records, Git history, and a formal counter-notice.

SEO + GEO Ripples

  • OpenAI Says Robots.txt May Not Apply To ChatGPT's Fetch Bot: About 15% of identified AI page-fetchers reached URLs European sites had marked as disallowed in TollBit's first-half 2026 data. OpenAI's documentation says ChatGPT-User may not follow robots.txt because a person initiated the request. OAI-SearchBot, not ChatGPT-User, controls whether a site can appear in ChatGPT search results, so blocking both trades away search visibility without fully controlling user-requested fetching.

  • Google Lowers Subscriber Requirements For Search Profiles: Google cut the threshold from 100,000 to 35,000 followers for YouTube, Instagram, and X. TikTok remains at 100,000. Google says Search Profiles give creators and publishers a shareable place in Search for their cross-platform content, and it plans to expand the feature to more countries.

  • OpenAI Makes The ChatGPT Sources Less Visible: Some ChatGPT responses now place “View sources” inside the three-dot More actions menu instead of showing a Sources button below the answer. The same citation panel remains available, but the extra click can make source discovery and referral traffic less likely.

  • Spammy Google Selected Canonical In Search Console: Search Console can sometimes show an unfamiliar or spammy domain as Google's selected canonical. John Mueller says shared interstitial or parked pages can cause this result. His advice is to check again after a few weeks, then post full details in the help community if it does not settle.

AI

Anthropic says future Claude models will use a version of Google DeepMind’s SynthID-Text method to mark generated text. The invisible pattern uses otherwise low-stakes word choices, adds no hidden characters or extra tokens, and carries no information about a user, organization, or chat. It can estimate whether Claude was involved in a passage, but it cannot establish authorship, identify another model, or reliably assess a short sample.

Google introduces Sheets canvas, a Gemini-powered read-write layer that turns a Google Sheet into an interactive mini-app from a natural-language prompt. Teams can edit the canvas or the underlying sheet, with changes syncing in real time, then share the canvas as a regular Sheet tab. It is available globally in English for Google AI Pro and Ultra subscribers, with a rollout starting for specified Business, Enterprise, and Education plans.

AI Ripples

  • Introducing Gemini 3.7 Flash: Google’s new coding and agent model scored 43.6% on FrontierCode 1.1 Main versus 34.4% for Gemini 3.6 Flash, and 65.3% versus 49.0% on DeepSWE v1.1. Google is pricing it through December 31, 2026 at $0.75 per million input tokens and $3.75 per million output tokens, before the listed price doubles on January 1, 2027.

  • Previewing Ultrafast mode: GPT‑5.6 Sol at up to 14X the speed: OpenAI is previewing an API tier for GPT-5.6 Sol that it says runs up to 14 times faster than Standard processing and produces up to 750 output tokens per second through Cerebras. Access is limited to an initial customer group while OpenAI studies uses in coding, commerce, research, support, and other time-sensitive workflows.

MARKETING

Erin Rodrigue separates positioning, messaging, and voice, then lays out a messaging framework built from purpose, value proposition, audience, message pillars, voice and tone, and proof points. Her practical test is useful: if a competitor’s name can replace yours in the copy without changing the meaning, the message is too generic.

My Take: This is why Floyi’s foundation is the brand and audience. A topical strategy only works when it starts with a clear view of who the business serves, the job that audience needs done, and the distinct proof behind the promise. Otherwise, AI can scale output, but it only scales generic copy faster. Watch a YouTube Short on Floyi’s approach to creating unique content.

Leigh Buttrey explains that Demand Gen moves PPC beyond capturing an existing search. The format needs images, video, audience signals, and creative that earns attention before a buyer has named the problem. Google says advertisers using both image and video assets saw 6% more conversions at the same spend than those using images alone, based on its own global campaign data.

LINK BUILDING + DIGITAL PR

Gaetano DiNardi says high-domain-rating placements, paid mentions, and link exchanges can create the appearance of authority without giving AI systems enough context to recommend a brand. He points to the usual failure pattern: an irrelevant insert, traffic from the wrong market, a page stuffed with unrelated outbound links, and a report that measures rows delivered instead of business impact. His alternative is a broader record of credible, topically relevant sources that connect the brand to its use cases and category.

SOCIAL MEDIA

X Open Source says X has released code affecting post ranking, filtering, and visibility in the For You timeline, alongside an Under the Hood pilot that shows eligible users labels applied to their accounts and posts. The X GitHub repository includes the ranking pipeline, visibility-filtering components, and documentation on what remains excluded to reduce gaming.

The YouTube Team says the Partner Program will change on February 1, 2027, with higher thresholds for Shorts revenue sharing and for new applicants seeking ads and Premium revenue.

  • Creators need 10 million qualified Shorts views in 90 days to earn ad and subscription revenue sharing from Shorts.

  • Channels below that mark remain in YPP and can still earn from long-form content. Shorts revenue sharing resumes if they cross the threshold again.

  • New applicants need 8,000 qualified watch hours in 365 days or 20 million qualified Shorts views in 90 days for ads and Premium revenue sharing.

  • Premium Lite is expanding to every country with YouTube Premium. It allocates 60% of net subscription revenue to its creator pool, compared with 30% for Premium.

  • YouTube reports more than 200 billion daily Shorts views and more than one billion hours of TV watch time each day.

Andrew Hutchinson reports that YouTube is rolling out Practice Mode for vertical livestreams. Creators can privately test chat, polls, stickers, green screens, and the stream setup, then go live from the same screen. YouTube is also rolling out a Get Feedback option in the Shorts composer that offers pre-publish notes on a clip's hook, visuals, and pacing.

X Creators says new enrollments in Revenue Sharing have closed, while existing members can keep earning through September 7, 2026. From September 8, eligible existing members can apply for the Original Content Rewards Program. Applicants need a Premium subscription, 500 verified followers, 500,000 verified-user Home Timeline impressions in the prior 90 days, and regular original content.

WHAT I'VE BUILT FOR YOU

Floyi - Build Topical Authority that wins in Google and AI Search. Don’t just plan your content strategy - make it unstoppable.

ReplyDeck - Be active on Reddit without living on Reddit. Schedule replies, stay present in the threads that matter, and get your time back.

TopicalMap.com Service - Let us do the heavy lifting. We handle the research, structure, and strategy. You get a custom topical map designed to boost authority and dominate your niche and industry.

Topical Maps Unlocked 2.0 - Unlock the blueprint to ranking success. Master the art of structuring content that search engines (and your audience) love - and watch your rankings soar.

AD

Build a Holiday Creator Affiliate Program in 90 Days

Creators lock in holiday content calendars 90 days out, before brands figure out commissions. Waiting too long to launch an affiliate program means less runway to build demand and a missed shot at the best partnerships.

The 90-Day Holiday Sprint covers commissions, recruiting, and scaling a program at Day 30, 60, and 90.

What Did You Think of This Week's Wave?

Login or Subscribe to participate in polls.

LIKE DIGITAL SURFER?

Find me and others in the Digital Surfer Discord community.

I’d also love to know what you think and if you have any ideas for the newsletter. Reply or email me at [email protected].

I’d also appreciate it if you shared it with fellow digital surfers.

You currently have 0 referrals, only 3 away from receiving LinkedIn Shout-out.

Have a great week taking your SEO and digital marketing to another level!

And don’t forget to drag the Digital Surfer emails to your Primary Inbox 🌊 

Reply

or to participate.