Profound Workflows is Profound’s automation layer for content operations in AI search. The Profound Workflows builder is click-based: marketing teams turn content audits, content refreshes, research-backed briefs and recurring reports into workflows that fire on a schedule or a trigger.
What is Profound Workflows?
Profound Workflows is a click-based automation layer, launched in public beta on December 11, 2025. Profound’s own description at launch was “a flexible automation layer built directly on top of our AEO platform” (AEO is answer engine optimization): teams click nodes together to codify a best practice into automated checks, analyses, insights and actions (launch announcement, December 2025).
What Profound Workflows automates well is the scheduled half of AI-search content operations; the judgment calls that decide citations are where it stops, and where an AI visibility platform earns its keep instead.
The name has shifted under it. In January 2026 Profound renamed the product line Agents, framing the move as the platform going from running manual processes to working alongside marketers, and the January updates landed for Enterprise customers first (release roundup, February 2026).
The current meaning is narrower. Agents sit inside Profound’s AI Marketer suite, called Aim, where you describe the job and Aim drafts the agent for you. The node-based builder survives inside Agents, still called Workflows, and a Migrate option converts an old Workflow into an Agent while leaving the original untouched (next-generation Agents post, October 2026).
So when someone says “Profound Workflows” in the product today, they mean the node-based builder inside Profound Agents. The launch-post meaning, the whole automation product line, is what most of the search results still describe, kind of frozen in time.
What Profound Workflows automates
| Workflow type | What it does | What comes out |
|---|---|---|
| Content audit | Checks pages for AEO, SEO, UX and brand alignment | An audit of what fails, page by page |
| Content refresh | Upgrades pages for AI-search performance | Refreshed pages, on a schedule or a trigger |
| Briefs and outlines | Generates research-backed briefs and outlines | A brief a writer can start from |
| Competitive analysis | Analyzes competitive insights in your category | An analysis of the pages ahead of you |
| Content-health reports | Reports on content health and gaps, ongoing | A recurring report |
| Recommendation triggers | Sends recommendations to writers, editors and subject-matter experts | Assigned work routed to a person |
Those six come straight from the launch post’s use cases. At launch you couldn’t schedule automatic runs, connect your existing tools or start from pre-built templates, and the launch post promised all three “by year’s end”.
What has shipped since is real, and Enterprise-shaped. The January release brought Profound data nodes, CMS integration nodes for WordPress and Sanity, a Slack node, a conditional node, version control and six pre-built agents, announced for Enterprise customers. You’ll find templates on the Agents page now, and a Migrate option lets you hand an old Workflow over to the newer describe-the-job agents.
How Profound Workflows works
The mechanism is a node graph. You codify a best practice as checks, analyses and actions, wire the nodes together, and the workflow runs when something tells it to. Profound data nodes pull the platform’s answer data in, so a branch can react to what ChatGPT or Perplexity actually said about you, and integration nodes push results out.
Triggers are schedules or events. The documented examples in their next-generation post, currently: your visibility on ChatGPT dropping more than 10%, a new page becoming a top cited source, a competitor publishing, a Slack message.
Connectors are MCPs, Model Context Protocol servers. Custom MCPs are supported from the Integrations page, and Slack, Gmail and Google Drive are the named ones.
There are two ways to build now, and the node-based builder is one of them. The other is Aim: you describe the job, and it drafts the agent in plain English.
Profound’s own line for choosing between them is “If you’d ask it in a conversation, use Aim. If you’d assign it as a job, make it an Agent.”
Workflows, in their words, “remain as a tool in your arsenal”, so nothing you’ve already built goes away.
How to set up a Profound workflow that holds up
A workflow that holds up starts from the number it should move, and every check in it has a threshold, a trigger and an owner.
1. Start from the number it should move
Name the metric the workflow exists to move before you build anything: citations per page, the share of answers that name you, the share that recommend you. Write it on the workflow itself. If you can’t name the number, you’ve built a report generator.
Our agents start one step further back, from the buyer question the work should win and the assistant it’s losing on. That way every run downstream is pointed at a number your executives already care about.
2. Write the check with a threshold and a trigger
“Monitor content health” is a mood, and a mood never fires on anything. A check has a threshold, a cadence and an owner: flag any page whose citations dropped more than 20% week over week, run it every Monday, route the flagged pages to whoever owns them. But miss any of the three and you get a list nobody reads.
Our agents run the check that actually decides retrieval: they score whether each passage on the page clears what retrievers measure, so “can a model lift this chunk” is answered for every flagged page.
3. Route every output to a person before anything ships
The workflow’s job is the draft, and a person’s job is the verdict. Someone reads what came out, accepts it or sends it back, and the feedback is what makes the next batch better. Auto-publishing removes the only step that was protecting quality.
This is how our processes run too: supervised, with a person reviewing the output, until a process has earned unattended runs.
4. Score editorial quality alongside the search signals
A pipeline that only checks SEO signals ships confident nonsense. The gates that hold up are editor passes, fact checks and, where the content warrants it, legal passes. Those are the checks a node graph can’t make, because they take judgment.
That’s the part our agents are built around: the editor pass is a gate in the pipeline, and its notes feed the next batch, so quality compounds instead of resetting every run.
5. Log what changed, when and why
Attribution is the part teams skip. Log every change with the page, the date and the target, and leave a few questions untouched as a comparison, so when the number moves you can show the lift was your work. A lift you can’t attribute is luck you’re taking credit for.
Our work log puts the dated record of every change on the same chart as the result, which is what makes the comparison readable in a review.
Where click-based workflows stop
Click-based builders are for the runs that don’t need judgment. The work that decides citations is mostly judgment, and a node graph can’t hold it. What to rewrite, what to cut, whether the brief is any good: those decisions are the product, and they don’t fit in a conditional node.
We’ve landed on the same split in our own retrofit work, moving production content jobs off node builders and onto agents: agentic execution wins where judgment compounds, and deterministic node graphs win where runs are judgment-light. So scheduled reporting and fixed-format checks belong in the graph, and the rewrites are the agent’s job.
Profound’s own line moved the same direction: a click-based builder in December 2025, and by October 2026 the flagship way to build is describing the job to an agent, with the node builder kept on as a tool. That sequence is public on their own pages. Our position is the stronger version: the judgment calls were never going to survive being encoded as nodes, and their right home is an agent running a documented method.
There’s a second cost, and it’s a little less visible. Everything you codify, the checks, the tuning, the prompts, lives inside the platform. We’re against any platform that tries to own your content: the alternative is agentic orchestration, your agent, your files, the knowledge plugged in over MCP.
And automation amplifies whatever method you already have. Scaled automated content that isn’t any good rises fast, then tumbles once the scale of the poor quality catches up with it. The schedule won’t save you from a weak method, and neither will the tool.
The refresh work is the receipt. In our 10x AI Visibility Refresh study we refreshed 98 posts, published nothing new, and citations went up 10x in 91 days (10x AI Visibility Refresh). The cadence mattered far less than the method each refresh followed.
Your agent is the workflow builder
The judgment half of this job is the half Omnipresence was built for. It plugs the AI SEO playbook into the agent your team already uses, over MCP: the audits, the rewrites, the refresh and the outreach, all run against the buyer questions you’re losing. Every run is supervised until it’s earned your trust, and it’s your agent and your files, so the method you build stays yours.
Frequently Asked Questions about Profound Workflows
How much does Profound cost?
Profound doesn’t publish standard pricing. It retired its Starter and Growth plans for brands in September 2026, and the $399 and $499 prices associated with those plans are no longer available for new subscriptions, per its pricing FAQ. Enterprise is custom, through a demo.
How is Profound Workflows different from n8n or Zapier?
n8n and Zapier are general-purpose builders: they route data between whatever apps you connect, and they know nothing about AI search. Profound’s builder lives inside its AEO platform, so the nodes read Profound’s answer data directly and the triggers are AI-search events like a visibility drop. The mechanics are the same graph of nodes either way, and the difference is what the nodes know.
Can Profound Workflows publish changes to my site?
Yes, through its CMS integration nodes: WordPress and Sanity are the named ones, and custom MCPs extend the connections to other systems. A Slack node routes drafts to people for review. Whoever reads the output before it goes live is still the quality gate.
Do you need Enterprise to use Profound Workflows?
The January 2026 updates were available for Enterprise customers, and self-serve users were directed to request a demo. That’s the last published word on tiers, and beta availability moves pretty fast, so check with Profound before you plan around it.
The scheduled half of this work is solved software. What decides citations is the method behind the run, and a person reading what came out of it.