The AI Citation Outreach Prompt I Run

Google Spam Update; Multimodal Search in GSC; Claude Breaks SEO; Google's SAFE Detects AI Spam; 35-Step Lead Gen Process; AI Citation Outreach; and Much More!

Author

Yoyao
September 26, 2026 • Est. Learning Time: 9 minutes

Sponsored by

FIRST …

Carlos Silva’s post on AI citation outreach (below) is a thorough playbook: a prompt-tracking export, three instruction files, a custom Claude Code command, and six opportunity categories.

I then thought, this is something that’s very simple due to Floyi.

If you’ve created a topical map in Floyi, your topics and AI citations are already there. One prompt in Claude/ChatGPT with Floyi’s MCP does the research and drafts the outreach pitches: 

The goal is to get our brand added to third-party pages that AI engines (Google AI Overviews, AI Mode, Gemini, ChatGPT) cite for our commercial-intent topics. When those cited pages mention or link to us, the engines are more likely to surface us. Find the cited pages, pick the ones that could realistically add us, find a real contact, and draft personalized pitches. Follow this process:

1. Pull the commercial-intent topics from our brand's map in Floyi, then pull the AI responses and citations for those topics.
2. Clean up the cited URLs by stripping tracking parameters and remove duplicates. Rank each URL by how many engines and how many of our topics cite it. Pages cited for several of our topics are the best targets, since one placement helps all of them. Priority topics are the ones where our brand is Not Present (neither mentioned nor cited).
3. Skip pages that won't add us: government and medical authorities, forums, Reddit, YouTube, marketplaces, direct competitors, and brands' own product pages. Also skip pages that never link out.
4. Sort the remaining pages into two tracks. ADD: we're absent from the page. FIX: we're mentioned, but the information is outdated, wrong or incomplete. Check FIX pages against our brand overview and knowledge base in Floyi, and quote the exact line that needs fixing.
5. Order both tracks from highest to lowest priority.
6. Find the contact the site publishes: author, editor or contact page. Don't guess email addresses. Draft a short pitch that names a specific section of their page and makes one clear ask. For ADD pages, give one reason our page helps their readers. For FIX pages, lead with the single correction. Only claim what our page and brand information actually contain.
7. Create a tracking spreadsheet of the target pages, tracks, contacts and pitches. Save the pitches as drafts. Do not send anything until I approve it.

One cut is on purpose. I skip competitors. Carlos keeps them in the pool. A direct competitor isn't adding you to their comparison page, and I'd rather spend outreach on pages that might say yes.

I kept the piece that matters most: pages that already mention you. If a cited page gets your pricing or features wrong, the engine can repeat it. Those are also the easiest pitches to land.

Have you done any AI citation outreach yet? If yes, what worked?

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

Google Search Central reports the September 2026 spam update started rolling out on September 24. The update applies globally and to all languages and Google says the rollout may take up to two weeks to complete. Site owners can follow progress on the ranking release history page until Google confirms completion.

Google reports Search Console now splits Web results into text-based and multimodal traffic. The new filter covers searches that start with an image through Lens, Circle to Search, photo uploads and Chrome image lookup. It is rolling out globally from September 24 with data visible from September 10 in Performance and Generative AI reports. Text query data is not shown for multimodal because the query is an image rather than words.

My Take: I tell every ecommerce and local client to treat product photos as search assets now. Keep clear images of your products on your site for multimodal search. Drop a photo into Google Search yourself and see if Google sees the same thing you expect shoppers to see or take pictures of.

The Foundation Team tested 20 ATS buying questions across six AI surfaces and found a clear versus layer deciding product comparisons.

  • 58.7% of surface-weighted source presence came from comparison pages and listicles.

  • 42.1% came from dedicated comparison pages on direct two-product questions.

  • Four vendors supplied only 27% of evidence on direct comparisons.

  • 42.7% came from independent publishers, 20% from other software vendors, 7.4% from review sites and 2.9% from community sources.

  • Listicles held 51.1% on neutral prompts versus 12.6% for comparison pages.

  • Comparison pages rose to 33.8% once a vendor was named and listicles fell to 21.4%.

  • Product pages and docs supplied 77.3% of ChatGPT inline sources with 63.8% from product pages alone.

  • Comparison pages and listicles supplied just 3.3% on ChatGPT versus 89.4% on Gemini and 72% on AI Overviews.

Barry Schwartz reports spammers indexed more than 12,000 public Gemini Notebook pages to push peptides, coupon codes, porn, and other junk. He says Gagan Ghotra flagged it as parasite SEO and Glenn Gabe reshared it after Charles Floate first showed how the pages could surface in AI Overviews. He notes Google left the pages live for days even though it could wipe the directory quickly. The trick relied on public NotebookLM pages being indexable by default.

Update: Google removed all the pages from its index the next day. No loophole survives long once it gets shared publicly.

Bryson Meunier breaks down 10 cases from a 41-site audit for the Chicago Independent Venue League and his own app build where Claude applied SEO best practices without enough context. Some of Claude’s SEO errors:

  • One venue was flagged for SPA JavaScript crawlability issues when it was a plain one-page HTML site.

  • One venue was told to claim its Google Business Profile even though the listing was already claimed.

  • All 41 owner one-sheets carried a no-sitemap-submitted note with no Search Console access to verify it.

  • 111 duplicate title tags from show names appended to default.aspx went unnoticed until he flagged them.

  • 78 link sheets lost suggested targets between sessions alongside fabricated contacts.

  • Canonicals pointing every demo page back to the demo homepage were reapplied 3 times after correction.

  • A full-app noindex was executed despite project instructions to keep demo pages indexed.

My Take: I’ve hit similar issues when I let Claude run SEO tasks on its own. It moves fast and sounds sure, and that is exactly when it slips in fixes I never asked for. Even when I give explicit instructions, Claude still feels like it should “do this” because it thinks it’s best practices. I keep it on a short leash now and check every change before it ships.

Lawrence Hitches explains why the invisible watermark on Claude output does not trigger a Google penalty. He notes every Claude model since August 2 biases word choice toward a secret half of the vocabulary so the mark travels with pasted text, and he points to the EU AI Act transparency code that took effect the same day. He says Google drew its line in February 2023 around scaled content abuse, meaning many pages with little value built to move rankings, and he cites the SEO heist case with roughly 1,800 articles and 3.6 million visits that later collapsed. He adds that a Semrush study of 42,000 posts found Position 1 pages were 80.5% likely human-written, which he reads as proof that thin work cannot hold the top rather than proof that AI cannot rank.

My Take: This is why we built the Brand Knowledge Base in Floyi. You can upload up to 2,000 of your own files and data points so briefs and articles start from your proof instead of generic text. Pages built on your numbers and your experience pass the useful and information gain tests. I also got this email from Claude about their watermarks for past models:

SEO + GEO Ripples

AI

Roger Montti reports Google published a three-page paper on SAFE, short for Scaled Abuse Forensics Examiner, built to catch AI-generated spam at scale. He notes the paper stays vague on test results but confirms the system is already deployed and uses four specialized agents coordinated by a root agent to review content, behavior and network links. He points out SAFE targets spirit of policy violations with a few-shot trained LLM, so it can flag content that breaks intent even when no existing classifier fires. He connects it to the September spam update and calls it Google's second AI slop system this year after S-CTS.

AI Ripples

  • GPT-6 Sol and Luna bring Astra-level intelligence at much lower cost, with API prices cut in half versus the 5.6 series. Sol targets complex coding work while Luna handles high-volume clerical tasks, and OpenAI says Sol makes about half as many factual mistakes as its predecessor. Both roll out now in ChatGPT Work, Codex and the API.

  • Claude Opus 5.5 matches Fable-level performance at 40% lower cost than Opus 5, with faster output and cheaper cache reads. It leads Anthropic's lineup on agentic coding, computer use and knowledge work and posted the strongest scores to date on its behavioral audit. It is live now on all platforms including AWS, Google Cloud and Azure.

  • Google Vids now generates free HD video with Gemini Omni for anyone with a Google account. New controls cover scene length, extensions and 1080p output, plus SynthID watermarking on every clip. Text-to-speech narration in 100-plus languages is coming soon.

  • OpenAI contractors fired for training AI with AI after 404 Media found multiple cases, with Glenn Gabe noting the setup mirrors Google with about 10,000 raters. It flags model collapse risk when new models learn from AI output instead of human judgment. Expect more human review layers as AI output keeps flooding training data.

  • Sam Altman urged global AI safety cooperation at the UN alongside Anthropic chief Dario Amodei on September 23. He called for human control and shared standards as systems grow more capable. The pitch landed a day after renewed debate over whether global oversight helps or slows builders.

MARKETING

Andy Crestodina breaks down lead generation into a 35-step checklist covering strategy, site build, copy, content, promotion, AI visibility and measurement. He focuses only on organic inbound and ties each step to clarity and trust signals that also shape AI recommendations. He notes that content programs with original research are 50% more likely to report strong results and that his team sees AI visitors convert at 3 times the rate of other visitors. He also flags that AI visits homepages 15 times more than other page types, so core answers need to live there and not in PDFs.

LINK BUILDING + DIGITAL PR

Carlos Silva walks through an AI citation outreach workflow that finds pages AI answers cite and gets your brand included on them. He cites Semrush research with Kevin Indig showing 62% of AI citations do not produce a brand mention. He lays out a six-step process using Semrush Prompt Tracking Sources export and Claude Code to classify pages as Not mentioned, Underrepresented, Outdated or Negative. He shows an H&M workwear example to stress human review before pitching and shares a low-lift pitch template where 8 of the last 14 replies agreed to take updates.

SOCIAL MEDIA

Andrew Hutchinson breaks down Adam Mosseri's latest algorithm videos and pulls out what creators can act on.

  • Ranking predicts actions like watching a clip or sharing it with a friend.

  • Recency carries heavy weight and the feed stays roughly chronological.

  • Original posts get precedence over reposts from other accounts.

  • Small public accounts get test distribution to non-followers to gauge response.

  • Users can steer the feed with favorites, Not interested flags and topic controls.

  • Creators should post as often as they can sustain and use trial Reels to test ideas.

Andrew Hutchinson breaks down advice from YouTube Creator Liaison Rene Ritchie and Global Content Strategy Lead Kyle Monahan on switching a channel focus without losing the audience. He says the first call is why you want to pivot and whether viewers come for you or for the topic. He notes that style and voice often carry more weight than the topic itself and points to YouTube strategic guide questions plus Google Trends for spotting openings. He adds that smaller niches with no clear voice can beat crowded verticals if you bring something distinct.

Social Ripples

  • Meta Muse checkout deal with Shopify lets Muse agents find products and check out with Shop Pay across Shopify stores. Amazon asked to be removed from Muse over its rules against automated shopping tools. Watch product feed quality because AI agents will reward clear listings and punish thin ones.

  • X pulls standalone XChat app from stores only months after its April launch with encrypted messaging. References to XChat were also removed from X source code with no formal announcement. It points back to the single everything-app bet instead of splitting chat into a second download.

  • LinkedIn replaces AI post rewrite with Post Proofreader for Premium members with inline change tracking they can accept or skip. The shift follows more than 1 million uses of its AI slop reporting option. Keep drafts human first and use the tool for polish, not for generating the idea.

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

Watch the full system run live on Sep 30. Walk away ready to do it yourself.

Most founders have LinkedIn traction with nothing to show for it in the CRM. On Sep 30, Maria Gharib (Mindstream) and Valerie Chapman (Ruth AI) walk through the exact system live.

From AI-assisted content creation to sequenced outreach to booked meeting. You'll leave with a process you can run the same day.

Eligible startups also get the LinkedIn-to-Leads Toolkit: ad credits, Apollo, Captions, and HubSpot's Prospecting Agent.

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.