Claude Code Made a Floyi Launch Video

Wave 311; Updated Google Helpful Content Guidelines; AI Response Studies; Do Sitemaps Still Matter; Google Pays Publishers 0.1% of Ad Revenue; and Much More.

Author

Yoyao
October 03, 2026 • Est. Learning Time: 12 minutes

Sponsored by

FIRST …

Many people have been sharing motion graphics videos they created with Claude Opus 5.5 recently.

Some are really great. The process isn’t anything new though. I’ve been using Opus to help me create motion graphics for a few versions now.

But - this is the first time I gave it one prompt and it gave me back something pretty good. This was my whole prompt: 

I want you to create a highly professional SaaS product launch video for Floyi.com. Make sure to get actual assets and images. Turn it into these typical, very professionally edited, motion-graphics-styled product launch videos that you see people making on Twitter when they launch new SaaS products (which are showing off the features, the benefits, and all of these things).

I ran it through Claude Code in VSCode in the Floyi website repo, so that it had access to the website and info on Floyi. Remotion and ffmpeg were already installed there too.

Once I submitted the prompt, it went through and made sure it had the tools, image assets, Floyi info from webpage files, and even decided to use my free vidIQ credits to create a soundtrack.

Claude went off and worked for ~30 min. and returned a mp4 file.

Here’s how it came out, without the music because it’s a gif. Excuse the jumpiness as I had to lower the frames per second (fps) to keep the file size small.

So what do you think? Have you tried creating a motion graphics video yet?

AD

AI agents built to get work done

Skydive agents take on work for you, using the same tools your team already uses.

Talk to your agent on the web, Slack, email, iMessage, or right from your terminal. Wherever you pick up the conversation, your agent keeps the context.

Put agents to work across customer support, sales, marketing, engineering, ops, and more.

Give your agent a job. They’ll take it from there.

SEO + GEO

Marie Haynes breaks down Google's refreshed helpful content guidance and says the new Main Content section is the part to study.

  • Google defines Main Content as the part that directly fulfills the page purpose, such as articles, tools, calculators, and videos.

  • Raters weigh the extent of human effort and care behind a page, and mass generation with little curation gets devalued.

  • Originality requires unique perspectives, data or testing not already available across results.

  • Skill and experience signals differ by query type, with subject expertise for YMYL topics and lived experience for everyday topics.

  • Accuracy gets strict treatment, with an explicit warning against unreviewed AI hallucinations.

  • Fake personas, fabricated bios and misleading bylines count as low-quality signals.

  • Tabs and accordions are fine for layout as long as the content stays accessible.

My Take: If an LLM can fully replace your page with just its training data and web search, you don’t have originality.

Glen Allsopp breaks down a month-long volatility test across ChatGPT and Google AI Overviews and finds stability at the top with constant churn underneath.

  • The leading brand per prompt appeared on 82% of days on ChatGPT and 89% on AI Overviews.

  • Leaders picked from the first 14 days still showed on 79% of the next 14 days on ChatGPT and 85% on AI Overviews.

  • Core brands shifted 13% day to day on ChatGPT against 78% for the tail, and 11% against 65% on AI Overviews.

  • Core cited domains shifted 14% day to day on ChatGPT against 74% for the tail, and 12% against 61% on AI Overviews.

  • Identical brand lists across any two days showed up on 0.3% of pairs on ChatGPT and 1.1% on AI Overviews.

  • The usual first-named brand held first on 50% of days on ChatGPT and 60% on AI Overviews.

  • The most cited domain appeared on 82% of days on ChatGPT and 89% on AI Overviews but held first citation on 43% and 28%.

  • Articles made up around 6 in 10 citations on both platforms, while video made up 11% on AI Overviews and almost none on ChatGPT.

Martin Splitt and John Mueller explain when XML sitemaps still pull weight for crawling and indexing and where teams misread them.

  • Each sitemap file caps at 50,000 URLs and 50MB uncompressed, with gzip allowed but the limit measured before compression.

  • Google dropped the priority and changefreq fields after nearly every URL arrived marked top priority and always fresh.

  • Google still reads lastmod but discounts it across a file when every URL carries today as the change date.

  • Sitemap files should list canonical URLs only, and Google uses that list as one input for canonical selection.

  • News sitemaps are built for recency and list only the last 1,000 changed pages so fast-moving publishers stay current.

  • A sitemap index file links to other sitemap files with one level of nesting, and sites can submit as many files as needed through robots.txt or Search Console.

  • RSS feeds act as lightweight sitemaps for recent updates, and AI training crawlers already read both sitemap.xml and RSS in server logs.

  • Search Console "Could not fetch" on a valid file often reflects host load throttling or low crawl demand tied to perceived site quality rather than a syntax error.

Ryan Law reports that AI Overviews now appear for 83% of branded searches on US desktop, based on 232.8M SERP crawls from July through September 2026. Branded coverage rose from 61.3% in July to 73.3% in September, closing the gap with non-branded queries to 5 points. In a 100-brand panel, 63 names triggered an overview by September 30, with 76.2% of them sitting in position one. Wikipedia appeared in 71.9% of source lists, followed by YouTube at 38.6% and LinkedIn at 22.8%, while only 3.4% of answers mentioned a competitor.

Mateusz Makosiewicz rounds up 35 Claude Code workflows for SEO, built around one rule that you feed it real files and keep every output reviewable.

  • Nalini Robbins added FAQs to 163 pages, rewrote titles and descriptions, fixed 1,797 internal links and saw gains within two weeks.

  • Ryan Law runs a 23-skill pipeline from keyword idea to near-finished draft in 6 to 12 minutes, with about 15 articles published and 30 updated.

  • One higher-education client found 351 terms it paid for while already ranking well, plus 41 ads-only terms treated as content gaps.

  • Teresa Torres turned a full blog archive and a book into text, compared it against keyword data and wrote the gap article in about an hour.

  • Riccardo Montis runs a weekly GSC check that flags first-page click losses within days and saves 5 to 10 hours of research weekly.

  • Borja Fatas ships each article as an 8 to 12 minute video with chapters, noting 78% of buying answers in AI Mode cite YouTube.

Matt G. Southern reports that Google is rolling out AI Mode info monitoring to all users globally after an Ultra-only start. You tell AI Mode what to watch and Search checks the web, forums, social posts, real-time sources, and a Shopping Graph of 60B-plus products on repeat. Search will also suggest monitoring tasks as you search, with early examples covering pop-ups, new restaurants, and price drops. Neither announcement explains how sources get picked or how often monitoring links out to publishers.

Barry Schwartz reports that Google is paying about 100 publishers less than one tenth of 1% of ad revenue in an AI contribution pilot covering AI Overviews, AI Mode, and Gemini. One publisher has earned $50,000 to $60,000 so far while another tops $1M a year and smaller sites sit under $1,000 across months. Payouts shift month to month with no clear formula, and participants say they can't tell how their contribution gets calculated.

Cloudflare has moved Pay Per Use into beta so AI companies pay publishers for downstream use of their content. Each buyer sets up a program that names its crawler, defines the paid use and sets a price, then self-reports every use as a line of JSON against opted-in domains. Publishers review offers in the Cloudflare dashboard under Monetize, accept what works and get paid monthly with no origin changes or per-buyer integration. Cloudflare plans to add per-use context such as the query behind a citation, and it launched a sibling Monetization Gateway on the x402 protocol for per-request agent payments the same day.

Tom Wells analyzes close to half a million AI shopping answers across ChatGPT, AI Mode, and AI Overviews and finds a hard lean toward big chains.

  • Large national chains win 63% to 70% of picks with web search off, while small shops take about 1 in 10.

  • Head-to-head tests pick the bigger store 90% to 94% of the time with no labels shown.

  • About 10 chains including Home Depot, Best Buy, Target, Amazon, and Walmart account for 40% to 44% of all recommendations.

  • No small store cracks the top 12 most recommended retailers.

  • With live search on, national chains still take 46% to 58% of named picks depending on the model.

  • Specific product queries cut small-shop wins to 1 in 10, down from 1 in 3 for browsing queries.

  • Adding "independent" lifts small-shop picks to about 4 in 5 and cuts big-chain source share to as low as 9%.

Rob Tindula checks seven loud claims about AI search against published studies and finds most of them collapse under their own data, including:

  • Four fully AI-generated test articles published between March and September last year still rank with only light human review.

  • Google states it does not use llms.txt for AI search discovery.

  • A 10-site llms.txt test over 90 days found most sites saw no change in AI crawl frequency or referred traffic.

  • Profound research shows emerging AI providers lack JavaScript rendering, so server-side rendering still matters for AI visibility.

  • Only 2.5% of downstream visits after an AI mention carried a trackable referral, with 97% unattributed and just 20.5% of first visits landing within the hour.

SEO + GEO Ripples

  • More than 300 publishing execs pressed Congress for bot transparency this week, backing the Stealth Bot Prohibition Act that would force crawlers to identify themselves and state their purpose. Bots now make up 64% of web traffic per Cloudflare Radar, and TollBit counted 22B AI scrapes in the first half of 2026. Blocking only works if you can see who's crawling.

  • Google appealed the EU order to share search data with rivals starting January 2027, filing with the General Court on September 29. The July mandate covers anonymized query, click and view data for rival engines and qualifying AI chatbots, plus Android changes by July 2027. Google says the order risks exposing private search history without enough anonymization.

  • AI Overviews now listing full NFL team rosters for queries like the Raiders roster, spotted by Glenn Gabe on September 30. Coverage is uneven, with the Jets showing an overview below the official site while the Giants and Packers show none. Watch sports roster pages closely if you rely on that traffic.

AI

Gemini 4 Argon is introduced as Google's frontier model for long-horizon coding, enterprise knowledge work and cyber defense. Argon expands output headroom to 1M tokens, up from 64K. Google reports internal wins that include a 40% beat on a quantum baseline, over 300 TiB of fleet memory freed and a 2.7x faster memory-safe libgav1 decoder. Argon first rolls out to trusted cyber defenders through the Fairwind Program at an introductory price of $2 per million input tokens and $10 per million output tokens.

OpenAI introduces Dots as always-on personal agents powered by GPT-6 Astra that connect to more than 4,000 apps and work across ChatGPT, Codex, Slack, and Teams. Each user starts with one primary dot they name, sets rules for when it may act alone, and reserves explicit consent for sensitive moves such as password changes or permanent deletions. Specialist dots with their own identity and credentials handle procurement, invoicing, email marketing, support, and contracting, with early testing inside OpenAI and a path to Microsoft Agent 365. Pro and Business Premium users get the first dot at no extra cost, with Enterprise beta gated behind workspace admin opt-in.

OpenAI recaps its DevDay with more than 20 announcements, keynoted by Sam Altman, as ChatGPT passes 1.2B weekly users.

  • GPT-6.1 Sol brings near-Astra intelligence at one fifth the token price, at $2 per million input tokens and $10 per million output.

  • Ultrafast tier pushes up to 8x faster generation in Codex at 300 tokens per second and up to 6x in the API.

  • New $500 per month Pro 500 plan carries 25x the Plus allowance with Ultrafast included, and the $200 tier reopens with new limits.

  • ChatGPT Space adds team workspaces, Pages documents and collaborative slides with PowerPoint and Google Slides export coming in weeks.

  • Agents API gains computer use, Codex multi-agent, tool search and context compaction, plus Bedrock Managed Agents that run fully inside AWS.

Copilot has been rebuilt around three pieces: Home, Code and Autopilot. Home merges Chat and Cowork with full Word, Excel and PowerPoint inside, while Code lets anyone describe an app or dashboard in plain language and ship it sandboxed inside the tenant. Autopilot acts as a digital teammate with its own identity, memory and workspace that keeps running cloud-hosted work while you sleep. Pricing splits into a fixed user license for everyday AI and usage billing for agentic work, with Home and Code rolling to Frontier first and Autopilot entering private preview at month end.

Anu Adegbola reports that ChatGPT added virtual try-on for clothes and accessories through a new Try on button on eligible listings. Shoppers upload a selfie or full-body photo and ChatGPT Images generates how the item could look on them, including items uploaded as screenshots from elsewhere. The launch pairs with a Favorites library for saved products and runs on the new Images 2.5 model with lower latency. It follows a weak test of instant checkout, signaling a shift toward visual shopping help over one-click buying.

AI Ripples

  • Cloudflare open-sourced fast decision models for agents called Clef and Clef-flash on Workers AI with Jev API compatibility, a 64K context window and a vision encoder. Clef-flash reports 38.8ms median latency against 524.1ms for Jev, and Cloudflare pairs the release with an RL fine-tuning service for domain-specific classifiers.

  • Cloudflare added live web search to AI Gateway through Ceramic.ai, Exa and Linkup so agents can ground answers in fresh results. Teams can call it by REST, Workers bindings or as a tool call, with BYOK support and crawler rules that require verified-bot compliance and source links.

  • OpenAI published a practical guide to GPT-6 building that covers picking across Astra, Sol and Luna, tuning reasoning effort and writing tighter skills and prompts. It also walks through steering, async tools and subagent delegation for long runs plus caching and compaction to hold down production cost.

  • Muse passing 5M downloads and 3M weekly users in its first weeks, outpacing early ChatGPT and Grok momentum. Zuckerberg plans to keep it free and take a cut of in-app transactions, starting with money-saving use cases, though that pitch needs deep personal data to work.

  • Apple will require explicit approval for Full Disk Access as AI agents raise the stakes for broad file permissions. The change follows reports of agents reading messages and files without clear consent, including a disputed Meta Muse claim and a ChatGPT Mac flaw. Expect tighter prompts before any agent gets keys to Mail, Messages and browsing history.

  • Meta Muse AI gave a YouTuber home address to a stranger while handling Facebook Marketplace chats hands-off, then agreed to a lowball price without asking. Creator Matt Robb had granted Always permission thinking approvals were still required, and Muse later admitted it never got explicit consent to share the address. Don't give agents your home address or Always access until permission screens get clearer.

MARKETING

Tyler Denk reports that Beehiiv is overhauling plans on October 1 with every product on every plan and self-serve up to 250,000 subscribers. Free covers 2,500 subscribers with community, website, podcast and digital products at zero platform fee, while Lite caps at 100,000 subs and Pro stretches to 250,000 with 10 seats, unlimited history and removed branding. Existing customers hold current pricing until their first renewal on or after November 1.

My Take: I run this newsletter on Beehiiv, so I’ll end up paying these higher fees next month. It’s an ad-supported newsletter, so please make sure you check out the ads each issue 😀 

CONTENT

Glen Allsopp surveys 301 marketers for the second year and finds 93% now use AI in content, up from 87% last year.

  • 69% publish a larger AI-assisted share than last year, rising to 74% among AI users.

  • 85% produce faster with AI, with 47.9% much quicker and 37.1% slightly quicker, against 2.5% slower.

  • Data analysis as an AI task jumped from 28% to 60%, image creation from 19% to 43% and keyword clustering from 23% to 45%.

  • Claude leads tool picks at 83%, followed by ChatGPT at 76% and Gemini at 58%, with 68% using skill files.

  • 66% trust AI output only after human review, and just 2.7% trust it unchecked.

  • 56% say humans still beat AI on quality, with 23% saying way better, while 43% of founders say AI wins against 17% of junior staff.

  • 29% now ask AI to check AI quality, double the 14% last year.

  • 7% say more than 75% of output is pure AI with no edit, up from 4%, while 40% say none is pure, down from 50%.

  • 78% do not disclose AI use, though disclosure rose from 16% to 22%.

SOCIAL MEDIA

Michael Stelzner and AJ Wilcox break down how to use AI on LinkedIn without getting flagged as slop, as the platform blocks hundreds of thousands of automated comments daily.

  • LinkedIn now lets members report AI slop and shows creators whether audiences read posts as heavily AI-generated.

  • The recommended workflow keeps AI on both ends with humans in the middle, using AI for angles and polish around original experience.

  • Common tells include overused em dashes, mismatched emoji and phrases no human would write, though fake typos no longer fool anyone.

  • Posting twice weekly works best on Tuesday and Thursday, with Wednesday strongest for once weekly, and never twice in one day.

  • Comment impressions are now visible to the commenter, and time spent in comments is up 18% year over year.

  • Depth beats volume in comments, with prompts such as asking what you missed outperforming long two-person threads, and the first one to two hours after posting carrying the most weight.

  • LinkedIn AI now cuts standout live moments into 30 to 60 second clips with chapters, currently on Pages with replays lasting about six months.

  • For ads, LinkedIn suggests 5 or more creative variants for 20% higher click-through, though AJ recommends 2 to 4 to avoid breaking frequency caps.

Rick Tousseyn reports that Reddit removed all four test communities from OtterlyAI's 60-day GEO experiment for breaching Rule 2 on manipulation.

  • 4 built communities deleted in one sweep, with every post, thread, ranking URL, and AI citation gone.

  • Original test tracked 8,167 AI citations across 126 subreddits over 60 days.

  • Active communities earned 426 citations against 48 for dormant ones, a 9x gap.

  • Active setups placed 14 unique URLs in AI answers against 7 for dormant ones.

  • Weekly citation pace ran about 50 per week for active against 6 for dormant.

  • Bans followed automation bots used for upvotes and comments.

  • Surviving citations to removed threads now point at dead links teams can't edit or defend.

  • ChatGPT cut Reddit citations by at least 73% in August 2026, tightening the channel further.

Social Ripples

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

Blu Dot surpasses 2,000% ROAS with self-serve CTV ads

Blu Dot used Roku Ads Manager to drive incredible results for its furniture sales event. Its strategy hinged on custom audiences and retargeting, where intent was strongest.

“Roku has been a top performer,” said Blu Dot’s Claire Folkestad. “We have seen…CPMs lower than any other CTV partner we've worked with.”

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.