The Visibility Report #11

Week of March 31 -- April 6, 2026

I'm feeling this one personally. Over the last several weeks I've gone all in on OpenClaw for both personal productivity and building agentic workflows.
​
And it was all built on a Claude Max subscription.
​
Anthropic had a rough week. Token quotas burning 10-20x faster than expected. Two thousand internal source files leaked in an npm package. And then, effective April 4, the company shut off subscription access for third-party agentic tools entirely. If your AI visibility workflow runs on Claude through something like OpenClaw or another harness, you're now paying API rates or switching models. The platform risk that AI SEO teams keep abstractly worrying about just became concrete.

Also this week: ChatGPT search is citing fewer domains per response (fewer winners sharing a smaller citation surface), 63% of US adults say ads in AI search would reduce trust right as self-serve opens up, and Grokipedia's Mt. AI crash confirms what we suspected -- drop in Google = drop in AI surfaces.

🔥 Anthropic's Tokenamageddon: Three Beats, One Warning for AI SEO Teams

Three separate Anthropic stories broke this week, and for AI SEO practitioners they add up to one thing: the infrastructure you're building your AI visibility workflow on is subject to platform risk you don't control.

Beat 1: The Token Drain Crisis

Claude Code users on Max plans started burning through monthly token quotas in 19 minutes instead of 5 hours -- a 10-20x overconsumption rate traced to prompt cache accounting bugs. Anthropic acknowledged the issue and shipped partial fixes, but the backlash continues. The users hit hardest were running automated, multi-step agents -- exactly the kind of pipelines AI SEO teams build for visibility monitoring, content optimization, and citation tracking.

📎 The Register | Forbes | DevOps.com | TheLetterTwo​

Beat 2: The Source Code Leak

Roughly 2,000 internal Claude Code files were accidentally shipped inside a public npm package -- exposed architecture diagrams, unreleased feature specs, and internal tooling. Anthropic called it "human error, not a breach." The leak gave the internet an unplanned look inside one of the most-used AI coding tools in the world. More relevant for AI SEO teams: it surfaced how Claude Code's agentic loops are actually structured, including the prompt patterns that drive automated workflows.

📎 Axios | The Guardian | VentureBeat​

Beat 3: The Third-Party Harness Ban

Effective April 4, Anthropic blocked Pro and Max subscription access for all third-party agentic tools -- tools like OpenClaw, anything running Claude through a wrapper outside Anthropic's own interfaces. Users relying on flat-rate subscriptions for automated Claude workflows now face API pricing: $15/$75 per million tokens for Opus. No warning. No grace period. If you were running AI visibility monitoring, content generation, or citation tracking through a Claude harness on a subscription plan, your cost structure just changed overnight.

📎 Dev.to​

Why it matters for AI SEO: The tools you use to track AI visibility -- the agents, harnesses, and automated workflows -- run on model infrastructure you don't own. Anthropic changed three things in one week: pricing behavior (token drain), security posture (source leak), and access policy (harness ban). Any one of those would have been notable. All three together is a signal. Evaluate your AI SEO tool stack's dependency on any single model provider.

📊 AI Visibility Research

ChatGPT Search Is Citing Fewer Sites -- and the Stakes Just Got Higher

Resoneo analyzed 27,000 ChatGPT responses over 14 weeks and found a 20% drop in unique domains cited per response after GPT-5.3 Instant -- down from 19 to 15. URLs per domain stayed flat at 1. That means the citation surface shrank, but concentration didn't. Fewer sites are splitting the same authority signal. If you're already in the citation pool, your share went up. If you're not, getting in just got harder.

📎 SEJ​

63% of US Adults Say Ads in AI Search Would Reduce Trust

An Ipsos survey of 1,085 US adults found nearly two-thirds say seeing ads in AI search results would reduce their trust in those results. The timing is notable: ChatGPT self-serve ads are now open. Google is testing ads in AI Mode. Early ChatGPT pilot CTR is running around 0.91% -- compared to Google's 6.4% for traditional search ads. The trust problem is real, but the platforms are pushing forward anyway.

📎 SEJ​

Conductor 2026 AEO/GEO Benchmark: AI Referral Traffic Is 1.08% -- and ChatGPT Owns It

Conductor's 2026 State of AEO/GEO report surveyed 250+ CMOs and analyzed AI referral traffic across 10 industries. The headline: AI-driven referrals now represent 1.08% of total web traffic. ChatGPT accounts for 87.4% of that. The absolute numbers are small. The concentration on a single platform -- one that just changed its subscription terms -- is the more important data point.

📎 Conductor​

🚀 Platform Updates

OpenAI Closes $122B Round at $852B Valuation -- Building the AI Super App

OpenAI closed a $122 billion funding round at an $852 billion valuation and announced plans for a "super app" bundling ChatGPT, Codex, browsing, and agents. Revenue is running at $2 billion per month. Self-serve ads are now open. For AI SEO practitioners: this is a company building a search, commerce, and agent platform -- not just a chatbot. Your brand's presence inside ChatGPT's ecosystem is going to matter more, not less.

📎 OpenAI​

Grokipedia's "Mt. AI" Confirms: Drop in Google = Drop in AI Surfaces

The Grokipedia case study is getting traction as the canonical cautionary tale for AI-generated content at scale. Glenn Gabe and Peec AI's analysis of its traffic pattern -- a sharp surge followed by a cliff drop -- confirms the "Mt. AI" shape: AI-generated content sites rise fast in both Google and AI search (AIOs, AI Mode, ChatGPT), then fall just as fast when Google's quality signals kick in. The correlation between Google ranking signals and AI citation signals is strong. Build for one, you build for the other. Lose one, you lose the other.

📎 SE Roundtable​

📝 From the Tool Blogs

Scrunch:7 Best AEO/GEO Tools for 2026 -- Competitive landscape roundup covering the full AI visibility tooling category, from upstarts to enterprise platforms.

seoClarity:ArcAI 3.0 Launch -- New suite moves from AI visibility tracking to actionable enterprise workflows, announced April 1, 2026.

Peec AI:Top Domains Cited by AI Search: 30M Sources -- Reddit, YouTube, LinkedIn, Wikipedia top the list across ChatGPT, Google AI Mode, Gemini, Perplexity, and AI Overviews.

Peec AI:Peec AI Referral Program Launches -- Active customers now get a referral link; 30% off for new signups, 20% revenue share for referrers.

Profound:Semrush Integration Nodes for Profound Agents -- Pull domain metrics, backlink profiles, and keyword research directly into Profound's AI visibility monitoring and content workflows.

AirOps:How to Build a Brand Kit That Makes Your Content Sound Like You -- Configure a single structured brand source of truth that any AI tool or MCP-connected workflow can reference for voice, tone, and content rules.

Semrush:What Is an AI Agent? (And What It Means for Brand Visibility) -- Primer on how AI agents research, evaluate, and act on behalf of users, and what that means for how brands show up in AI-mediated discovery.

💡 Strategy & Frameworks

The Mt. AI Pattern: Why AI-Generated Content at Scale Fails

Grokipedia is the data. The analysis is worth sitting with. AI-generated content sites get a visibility surge because they rapidly expand topical coverage -- AI search surfaces them because they appear comprehensive. Then Google's quality signals identify the pattern and suppress them. And because Google signals and AI citation signals are correlated, the suppression hits both channels simultaneously. The lesson isn't "don't use AI." It's "don't build visibility on content that doesn't actually satisfy the query." Volume without quality is a time bomb with a short fuse.

Platform Risk Is Now a Real Line Item in Your AI SEO Budget

Anthropic's week should go in the "case study" file. Three separate events -- a billing bug, a leak, and a policy change -- disrupted users in different ways but shared one underlying cause: dependency on a single AI provider with opaque policies. AI SEO teams that have built automated monitoring, content workflows, or citation tracking on top of a single model provider are running the same risk. Diversify the model layer or build abstraction into your stack. This will happen again.

🔮 What to Watch

1. ChatGPT citation surface narrowing. Fifteen unique domains per response instead of nineteen means higher stakes for every site in the pool -- and higher barriers for the sites trying to break in. Watch whether GPT-5.3 Instant's pattern holds as it becomes the default.

2. AI search ad monetization vs. trust. 63% of US adults say ads reduce trust. Both ChatGPT and Google are pushing ahead anyway. The first wave of brand safety incidents from AI ad targeting will reshape this conversation fast.

3. Model provider platform risk for AI SEO tools. Anthropic's moves this week are a preview. Any AI SEO tool that runs on a single model provider's API is subject to the same risk -- pricing changes, policy shifts, access restrictions. Evaluate your stack's provider dependencies now, not after the next disruption.

The Visibility Report #11 | Will Scott
This newsletter is produced collaboratively by Will Scott and Bob, an AI agent. Human oversight, AI efficiency.