The core distinction
Traditional GEO tools answer one question: are you cited? Agentic SEO answers a second, harder question: what did the tool actually do about it? Most of the category, general AI-visibility trackers included, stops at measurement and recommendation. Agentic SEO is the tier above: real, verifiable action taken on a site, not a report a human still has to implement somewhere else.
The five-tier taxonomy
This is a deliberately honest framework, the top tier is real, but it isn't "fully autonomous," and no one credibly claiming that today should be trusted with a client's live site.
Why Tier 4, not a false Tier 5 claim: the honest, defensible, sellable frontier isn't "we're autonomous", it's "we actually execute, with the control a professional needs." That's the wedge.
Where the market actually stands
This isn't a hunch, we ran 19 well-known SaaS brands through Citerank's own scoring engine to see where recognizable companies actually land on this taxonomy.
- Average score: 44/100, a D. The single highest score across all 19 brands was Zapier at 69, a C. Not one brand reached a B or an A.
- 11 of 19 sites scored a flat zero on agentic protocol support, nothing on the site is callable by an AI agent at all. That's not a Tier 4 gap, it's a Tier 1 gap: most recognizable brands haven't even started on Pillar B.
- Every brand in the study measures its own visibility in some form. None of them execute a fix directly, and none of them expose agent-callable actions on their own site.
That's the actual state of the category right now: almost the entire market is clustered in Tiers 1-3, on both pillars. The gap between "everyone measures" and "almost no one executes or prepares for agents" is the opening this framework describes.
The two pillars
Most of the category conversation collapses "agentic" into one thing, a tool taking action for you. That's only half of it. The other half is preparing your site to be found, understood, and used by every other AI agent that visits it.
Agentic execution
The tool acts on your behalf.
- 5 agent workflows · Citation Recovery, Page Optimizer, Competitor Leapfrog, Weekly Brief, Content: Research→Article→Deploy, each a real probe → synthesize → act pipeline
- Real WordPress execution via Citerank Deploy: schema deployed live, articles created as real drafts, both revertible. On sites that are not WordPress, a Cloudflare Worker deploys llms.txt, robots rules, WebMCP and A2A files
Agent readiness
Your site becomes something an AI agent can parse, trust, and act on.
- AX Agent Audit & AX Tree Compare, tests whether an agent can actually navigate a site, not just read it
- OKF Bundle Generator, converts a site into Open Knowledge Format, the structure agents consume directly
- WebMCP Generator, makes site actions directly callable by AI agents via MCP
- Entity Map, llms.txt, a2a.json, Speakable schema, all deployable live, not just downloadable
Most platforms in this category have no Pillar B: they measure citation and stop there. In Citerank, "agentic" means both "acts for you" and "prepares your site for every other agent that touches it."
What this looks like in practice
A concrete example, not an abstraction: run Page Optimizer on a live page, and the tool doesn't just tell you the FAQ schema is missing, it generates the JSON-LD and, if the page is on a connected WordPress site, pushes it live, with a revert option waiting if anything looks wrong. Run the Content agent on a topic, and it doesn't hand back a Google Doc, it researches what AI engines currently cite for that query, writes the article section by section, and creates it as a real WordPress draft with schema pre-attached. That's the difference between a report and an action, made concrete instead of asserted.