Three signals landed in the same week in April 2026. Individually, each is a curiosity. Together, they describe a shift most CMOs are not yet measuring.
Cloudflare reported that only 4% of websites are “agent-ready.” Buffer shipped /pricing.md, a plain-text pricing page written for procurement agents rather than buyers. And a piece in The State of Brand documented what marketers had been whispering about for months: LinkedIn’s ranking engine now quietly punishes company pages and rewards individual employees.
Three different surfaces. Three different vendors. One underlying pattern.
The audience that reads your marketing has changed. The measurement framework that tells you whether your marketing works has not.
The signal marketing teams keep missing
When we wrote Your Docs Have Two Audiences Now. One of Them Counts Tokens., the frame was engineering documentation. Addy Osmani’s Agentic Engine Optimization stack was, in its first form, a playbook for developer experience leaders. Pricing pages, brand content, and social reach were not in scope.
They are now.
The pattern that showed up in docs is showing up everywhere agents read. Pricing, expertise, trust, brand reach. All of them are becoming structural surfaces that agents parse on behalf of humans. The human still decides. But the agent is increasingly the one that shortlists.
Buffer’s move to publish a canonical /pricing.md is the loudest example. Rob Litterst at PricingSaaS framed it with the line worth memorizing: “Machine-readable pricing is a forcing function for pricing transparency and clarity.” You cannot write pricing for an agent and remain vague. The page either has numbers, tiers, limits, and contract terms, or it does not. The agent does not read between the lines. There are no lines.
This is governance by grammar. The machine-readable format forces a marketing function to commit to claims that the brochure format allowed it to hedge.
The 4% number is the whole story
Cloudflare’s Agent Readiness Score, a crawl-based assessment of how consumable a site is for modern agents, found that only 4% of sites met the bar in its first survey. The remaining 96% fail in ways that are not mysterious: missing llms.txt, missing structured metadata, blocked access for legitimate agents, token-bloated pages, untagged content, and authentication walls that treat an agent request like an attack.
Self-reported adoption of llms.txt has crossed 800 sites, including Cloudflare itself, Vercel, and Coinbase. Treat that number as directional rather than precise. It is a vendor count of a voluntary standard. The shape is still clear. The early adopters are a small, deliberate cohort. The long tail has not started.
Which means there is a window. The competitive surface of being legible to agents is not yet saturated. The companies that invest in it this quarter will have an 18-month head start on the ones that wait for a category analyst to declare it a category.
LinkedIn and the return of the person
The third signal is the most uncomfortable, because it reorganizes how B2B marketing has worked for a decade.
The State of Brand piece documents an observable pattern in LinkedIn’s ranking behavior since late 2025: company-page posts reach progressively smaller audiences, while posts from individual employees routinely outperform them by an order of magnitude. The article attributes this to 360Brew, LinkedIn’s unified 150-billion-parameter ranking engine launched in late 2025. The specific mechanism is not publicly disclosed. Treat the algorithm attribution as inference drawn from observed ranking, not confirmed internals.
But the outcome is not inferred. It is measured. LinkedIn is now the #2 most-cited domain across ChatGPT Search, Google AI Mode, and Perplexity in the April 2026 citation snapshot. Citation rank moves month to month, so do not read this as a permanent fixture. Read it as a signal of where trust has started to pool.
Two statistics from the same research deserve to be read together:
- 3% of employees share company content, and those 3% drive roughly 30% of brand engagement.
- 64% of hidden decision-makers, the finance, legal, compliance, and procurement layer behind the visible buyer, report trusting thought leadership content over marketing materials.
Both numbers come from a single source within the article, so carry them as directional rather than canonical. But the direction matches what every B2B marketing leader already suspects. Brand reach is consolidating around people, not logos. The agent-mediated discovery layer amplifies that, because personal expertise is easier to parse as a trust signal than corporate messaging.
The uncomfortable implication: the corporate LinkedIn page is a governance asset that has lost its distribution. The distribution now lives on the employee profiles that the marketing function has historically treated as “nice to have” rather than infrastructure.
Why this is a CMO problem, not a content problem
The instinct inside most marketing teams will be to treat all of this as a content tactic. Add llms.txt. Write a pricing.md. Start an employee advocacy program. Check the boxes.
That response misreads the nature of the shift.
Machine-readability is not a format. It is a governance posture. It requires the CMO to decide, deliberately and not by drift, what agents are allowed to see, what they are allowed to quote, what claims the brand is willing to commit to in a format that removes the marketing cushion, and who speaks for the company when the company page has quietly lost its voice.
As we argued in Advertising Discovers Governance. Two Years Late., the operational muscle for this work does not exist yet inside most marketing organizations. Engineering built policy-and-decisioning layers over years of painful incidents. Marketing has not lived through the equivalent incidents at the same scale. The conversational-AI placement work at OpenAI is starting to deliver them.
The CMO now inherits four governance questions that did not exist on the job description two years ago:
- What is machine-readable about our company, and who signs off on it? Pricing, product capabilities, SLAs, case study claims, team expertise. Each becomes an agent-consumed asset the moment it appears in a structured format. If nobody owns the canonical version, agents will synthesize one from whatever they find, and you will not like the result.
- Who represents the firm when the firm’s own page does not rank? If 3% of employees drive 30% of reach, employee voice is not a content channel. It is brand distribution infrastructure. That requires editorial standards, training, and governance.
- What is our citation posture? When ChatGPT, Perplexity, and Google AI Mode answer a prospect’s question about your category, what do they cite? The right measurement is not traffic. It is citation share. Most dashboards do not track it. Build the dashboard.
- Where are the agent-hostile surfaces in our own stack? The 25K-token page. The locked-behind-auth datasheet. The pricing page that requires a sales call. Each of these is a machine-readability failure. Each has a specific owner. The CMO’s job is to know which.
The new KPI
Traffic, impressions, and engagement assume a human on the other end of the session. Those metrics still matter. Humans still buy things. But they no longer describe the full funnel. A prospect who asked Claude or ChatGPT which vendors to shortlist, and who never visited your site, never appears in your dashboard. They appear in the competitor’s pipeline.
The KPI that captures this is not yet standardized. But its shape is visible. Machine-readability is the closest current proxy: a composite of agent-readiness score, citation share across major LLM answer surfaces, employee-driven reach as a percentage of total brand reach, and canonical-asset coverage for pricing, product, and expertise.
It is an uncomfortable metric because it rewards work that does not look like marketing. No campaigns. No creative. No brand film. Just the disciplined exposure of what the company actually is, in a format that an agent cannot misread.
That is the trade. You give up some of the ambiguity that made marketing feel like marketing. You gain legibility in a discovery layer that has already changed how buyers choose.
The CMOs who see this clearly will stop asking how to optimize their content for search engines. They will start asking who, inside the company, owns the fact that the company is now being read by machines.
The question will become a job description. Then an org chart. Then a budget line. In that order.
You do not need to wait for the title to exist. You need to start measuring what the title will eventually measure. That is the KPI.
This analysis synthesizes PricingSaaS’s How to Optimize Your Pricing Page for Agents (April 2026), The State of Brand’s LinkedIn Killed Company Pages (April 2026), and Cloudflare’s Agent Readiness Score (April 2026).
Victorino Group helps marketing leaders measure and govern machine-readability as a KPI. Let’s talk.
All articles on The Thinking Wire are written with the assistance of Anthropic's Opus LLM. Each piece goes through multi-agent research to verify facts and surface contradictions, followed by human review and approval before publication. If you find any inaccurate information or wish to contact our editorial team, please reach out at editorial@victorinollc.com . About The Thinking Wire →
If this resonates, let's talk
We help companies implement AI without losing control.
Schedule a Conversation