Research & InsightsBrief B-995
13 of 28 CMS Vendors Ditched Their Category Label. Documentation Tells You Why
Thirteen of 28 tracked CMS vendors changed category since 2023, and DXP use fell from 12 to 6. Agent identity, not branding, now separates them. Here is the six-question test for buyers.
Thirteen of 28 CMS vendors tracked since October 2023 have changed their category label, and DXP self-descriptions fell from 12 vendors to six. That scatter is the story. It also means the label itself has stopped working as a comparison tool, and buyers need a harder test.
The research, published Oct. 22, 2026, tracked homepage positioning across 28 content and experience vendors using the category word in each page's title, description or headline. AI-led or "agentic" wording rose from one vendor to 10 over the three-year window. The author coded the categories personally, so treat the tally as a judgment call.
Who dropped the DXP label?
Acquia relaunched on Sept. 14 as "the Agentic Content Platform." Five weeks later, its page title still read "Digital Experience Platform for Content, Community, Commerce." It was still unchanged on Nov. 13. Contentstack and Optimizely dropped the DXP label between June and September. All three cited governance in their announcements.
Four vendors — CoreMedia, HCLSoftware, Kentico and Liferay — kept "DXP" untouched for three years. At least five, including Kontent.ai and Optimizely, use "Agentic CMS" as a product label further down the page; none puts it in a title or headline. Vendors now scatter across five labels, and none dominates.
Sitecore CEO Eric Stine told CMSWire at the company's Orlando customer conference that he is not focused on labels and will leave that question to analysts. He said he is focused on customers.
What did the pages say before the announcements?
The archived captures undercut the launch-day framing. Contentstack's page title already read "Agentic Experience Platform" in April, two months before its June launch. Optimizely's homepage called itself an "AI-powered, all-in-one platform for marketing" in April 2025, well over a year before its Opticon conference at the end of August.
The governance promises themselves are quotable and specific. Contentstack's June 9 release said agents without both context and governance "are just expensive guesses." Optimizely's Kevin Li, SVP of Product, said in the Virtual Teammates release that even when AI does its part, "you don't always know who did what," and that every Virtual Teammate has "their own identity and permissions." Acquia founder Dries Buytaert made the same governance case on his blog.
Here's the skeptic's caveat. Each of those promises can be tested against documentation, and the documentation doesn't uniformly back the branding.
Colleague or proxy: which is the agent?
The review covered 10 vendors — including Contentstack, Optimizely, Acquia, Sitecore and Adobe — using public pages only, no trials, no logins. On one design choice they split cleanly.
Three treat the agent as a colleague with its own account:
- Optimizely: each Virtual Teammate "has its own Opti ID account" and acts as itself inside Optimizely's products with "only the access you grant it." It also "keeps its own usage separate from human usage."
- Acquia: the agent holds a dedicated Agents Team role that "never inherits permissions from the requesting user." Caveat: on Acquia's own Source CMS, the agent holds a fixed member role, so the real control is which sites it can reach.
- Kontent.ai: expert agents get their own permissions model.
Six treat the agent as a proxy for the person who started it:
- Contentstack: an on-demand run "acts as the person who ran it," unless the author fixed an account.
- Sitecore: its MCP server uses access that "follows existing Sitecore permissions."
- Adobe: Experience Manager and Coworker agents work with "permissions inherited from your Adobe organization."
- Storyblok, Hygraph and Contentful do the same. One vendor of the 10 could not be read.
The branding doesn't predict the choice. Kontent.ai kept its CMS label and sits with the colleagues. Contentstack and Sitecore both moved away from "DXP" and sit with the proxies.
What does the audit log actually show?
"Who did what" is a records question, and records live in documentation. Here the answers thin out fast. Acquia's documentation states "No customer-visible record of agent activity is provided" for its Work mode in the current release. Contentstack keeps a separate execution log of agent steps, but its content audit log does not label agent writes as the agent. Sitecore and Adobe state that audit logging exists, and no page in the review says agent actions are labeled as agents.
The stakes are concrete. Consider a content manager who asks an agent to rewrite 50 product pages before a campaign launch. If the agent runs on her permissions, the log shows her name on every page. She approves the batch, so maker and checker are one person — defeating the four-eyes principle (the maker of a change should not also approve it). Three weeks later, a page carries a claim legal never cleared, and the record can't say who wrote or approved it.
Nine of the 10 vendors document a person approving agent output before it goes live. That's a real control. But in six of those nine, the agent works with the starter's permissions, so the requester is often the approver. Human approval only works when the approver isn't the requester.
What should buyers ask before the next demo?
Forrester analyst Joe Cicman told CMSWire in April that "Agents are a forcing function on the operating model." Six questions test that — all answerable from documentation, not demos:
- Does the agent have its own account and permissions, or does it run as the person who started it?
- When the agent changes content, does the audit log name the agent?
- Who approves the agent's output, and can that be someone other than the requester?
- Which actions need approval, and who sets that?
- What happens when an agent goes wrong mid-run?
- What can be undone, and where is the record?
Only Optimizely documents a stop mechanism — deviation thresholds on its specialized agents — making it the least-documented capability in the set. Optimizely answers questions 1, 2 and 5 in writing; Acquia answers 1 plus a separate approval step before publication; Kontent.ai answers 2 and 6.
If most answers come back as "we'd have to check," get them in writing before the contract is signed. Several of these agents are weeks or months old and documentation moves, so ask for the page again at signing.
Source: feeds2.feedburner.com
Filed under
- cms
- dxp
- ai-agents
- content-governance
- vendor-comparison
More from Nathan Brooks
Show full bio
Staff writer covering companies, deals and strategy in marketing leadership at CMO Observer.
58 articles
The newsroom digest
One email a day with the stories that matter. No noise.




