Adobe just bought the dashboard. The work is upstream.

Adobe has announced it will acquire Semrush for $1.9 billion in cash. Twelve dollars a share, all cash, expected to close in the first half of 2026 subject to the usual approvals. Semrush slots into Adobe Experience Cloud alongside Adobe Experience Manager, Adobe Analytics, and Adobe Brand Concierge.

If you have been watching this space, the price tag matters less than the framing.

The sentence that re-prices the category

Adobe's stated objective is a complete solution that gives marketers a complete understanding of how their brands appear across owned channels, LLMs, traditional search and the wider web.

Read that sentence twice. The order is doing work. Owned channels first, the things you control. LLMs second, things you do not control but increasingly cannot ignore. Traditional search third, the thing the SEO industry has been working on for two decades. The wider web fourth.

Anil Chakravarthy, who runs Adobe's Digital Experience business, put it more directly: "We're unlocking GEO for marketers as a new growth channel alongside their SEO." Bill Wagner, Semrush's CEO, named the customer concern: "With the advent of LLMs and AI-driven search, brands need to understand where and how their customers are engaging in these new channels."

This is the moment generative engine optimization crosses from a niche topic discussed at SEO conferences into an enterprise budget line. The largest customer experience vendor on earth has just spent nearly two billion dollars to say so.

What the dashboard tells you, and what it cannot

A measurement platform is, by construction, a rear-view mirror. Semrush has built a good one for AI answer visibility: which LLMs surface your brand when asked about your category, what they say, how often, in what context. That is genuinely useful. Most enterprise marketing teams have been flying blind on that question for at least two years.

But measurement tells people something like: "your brand appears in twelve percent of relevant LLM answers in your category."

It does not tell them what to publish, in what shape, with what governance metadata, so that the figure becomes forty.

That is upstream work. It happens at the carrier layer, the source documents that LLMs and AI agents read before they form an answer. It happens in the structured data, the descriptive metadata, the licensing signals, the agent-readable instructions on each page. Once a brand's content has been indexed and inferred over, no dashboard can retrofit clarity that was not there at publication.

The dashboard is downstream of the decision that determines its reading.

Where established standards leave a gap

The web has been here before. SEO did not invent visibility; it operationalised standards that already existed: HTML, sitemaps, robots.txt, structured data, canonical links. Accessibility did not invent inclusion; it operationalised WCAG. Each of those movements succeeded because it sat on top of a standard, not in front of it.

The same is true now. Schema.org tells you how to describe a product. WCAG tells you how to make a page accessible. llms.txt and robots.txt tell crawlers and AI agents what they may and may not consume. sitemap.xml tells them what exists. Each of these is well-defined, widely deployed, and not in dispute.

What has been missing is the governance layer for AI and agent traffic specifically. Who is allowed to read this content. On what terms. With what attribution. With what verification that the document is current and from the named source. These are not questions the existing standards answer, because they were not designed to.

That is the layer Machine Experience operates on. It does not replace Schema.org or WCAG or llms.txt or sitemap.xml. It adds the small set of governance fields where they leave gaps. A well-built MX page is also a well-built SEO page, an accessible page, and a GEO-ready page. The economic argument for caring about that just got a $1.9bn floor under it.

What changes for anyone, or anything, that publishes

The audience is wider than most framing of this question assumes.

This is a problem for anyone, or anything, that publishes, rather than just authors. Everything a business puts on the web is now being read by machines before, during, or instead of humans: product pages, service descriptions, pricing tables, policy documents, API specifications. How that content is interpreted, summarized, cited, or acted upon is no longer a theoretical question; it is happening now, at scale, with or without the publisher's knowledge.

The agent web is not neutral infrastructure. Cloudflare blocks agents at the edge. Markdown-for-Agents proxies serve stripped versions of pages, with <meta> fields, structured data, and governance signals removed entirely. Answer engines summarize and drop attribution. An organization that has not expressed its content policy in machine-readable form is publishing without a contract into a network that will apply its own terms by default. Those terms are not the organization's terms. They are not the product team's terms. They are the default assumptions of whatever system ingested the content first.

Machine Experience governance fields are contract terms, not markup ornaments. mx:content-policy: extract-with-attribution is a machine-readable instruction that travels with the document. mx:status: current tells a summarizer whether the pricing is live or superseded. mx:origin gives the content a traceable source so that when an agent cites a product specification, the citation points somewhere real. mx:content-scope defines what the document covers so an agent does not generalize a returns policy into a brand-wide commitment. Without these fields, a product page that passes through a Markdown proxy arrives at an agent stripped of its provenance, its permission boundary, and its scope. It is no longer a document belonging to a business with stated terms. It is text that was found.

What changes is this: every entity that publishes to the web now publishes into a machine-read network. The governance layer is the minimum viable contract for any business that wants its content to represent it accurately when machines are the readers. Adobe's acquisition signals that the largest vendors have understood this. The argument for doing the work just received a $1.9bn floor. The gap was always there. The machines arrived and made it visible.

I have been working on this for two years. Drafting the standard, writing the books, building the audit tools, sitting in front of people who needed the fifteen-minute preamble before the conversation could begin. That preamble is now redundant. The largest customer experience vendor on earth has just delivered it on my behalf, in a press release, with a $1.9bn signature at the bottom. If you have been waiting for a moment to take this seriously, this is it.

And the law just arrived alongside the dashboard

The Adobe acquisition is one half of the story. The other half is the European Accessibility Act, Directive (EU) 2019/882. It came into force across the European Union on 28 June 2025. Public-facing PDFs, e-books, banking applications, ticket machines, and digital content from in-scope businesses must now meet the relevant accessibility standard, which for PDFs is ISO 14289-1 (PDF/UA). The penalties are real and the enforcement window has opened.

Note: This page describes regulatory frameworks in general terms only. Nothing here is legal advice. Requirements vary by jurisdiction, organisation type, and use case. Consult qualified legal specialists for guidance specific to your situation.

The law was written for human disability accommodation. The artefact it produces, by happy convergence, is the same artefact a machine reader needs. A tagged PDF carries a structure tree of headings, paragraphs, lists, tables, figures, captions, and reading order. A screen-reader user navigates that tree to skim the document. An AI agent ingesting the document reads the same tree, locates sections by heading level, walks tables row by row knowing which cell is a header and which is data, pairs figures with captions. The cognitive work that the screen reader does for the human and the cognitive work that the agent does for the machine are the same work, performed against the same metadata, producing the same correct answer.

An untagged PDF, the kind most public organizations have been shipping for thirty years, is a wall of positioned glyphs. Agents fall back to optical-character-reconstruction style guesswork: rasterise, classify, segment, hope the table grid recovers, hope the reading order doesn't leak across columns. The reconstruction is expensive and frequently wrong. The agent then quotes its made-up numbers as fact. The user reading the answer cannot see the reconstruction step.

So the upstream argument extends: machines need governance metadata to act correctly on web pages, and they need structural metadata to act correctly on every other carrier the publisher ships. The law has now made that structural metadata mandatory in the carriers that matter most.

This is where the audit work that this consultancy actually does for clients sits. The Web Audit Suite reads a published site and tells the publisher exactly where the machine-reader signal is missing: which pages lack governance metadata, which PDFs lack structure trees, which Schema.org claims contradict on-page text, which agents are blocked at the edge, which content-negotiation defaults strip MX fields in transit. The output is a list of specific, fixable defects with the page-level severity that an engineering team can prioritize. The same automated check that gates our own deploys runs against the client's site as a one-shot service.

Adobe just told the boardroom that this work is worth $1.9bn. The European Union just told the legal department that some of it is now mandatory. The audit is what bridges the two messages: it tells the engineering team exactly which lines need to change, in which file, by when. The work is upstream of the dashboard. It is also upstream of the EAA enforcement letter that some unprepared businesses will receive in 2026 and 2027.