This is an example of an agent you can build with Sylvie: give it a page, a draft or a target keyword and it returns a ranked set of H1s, title tags and social hooks, each one built for real search intent and sized to survive the results page. It runs on your brand brain, so the angles come from your positioning, your tone rules and the library you have already published, not from whatever is currently sitting at the top of page one. Sylvie does not ship it pre-built; your team assembles it on the brain that knows what this brand is trying to own.
Sylvie does not ship this agent pre-built. It is an example of an agent your team can build with Sylvie, powered by your brand brain, so it runs your workflow the way you actually run it.
The Headlines agent is an example of what a marketing team or agency can build on Sylvie: a headline engine for search, covering the H1 on the page, the title tag in the results, the meta description underneath it and the hook that carries the same piece into social. You can point it at one page or at an entire quarter of planned content.
Headlines are where SEO and brand voice usually fight each other. The version that nails the keyword reads like it was assembled by a machine, and the version the brand loves says nothing a search engine can place. An agent built on your brand brain ends that argument, because it holds the keyword target, the positioning and the tone at the same moment and writes a line that satisfies all three.
Before it writes an option, the agent reads the brand brain: the positioning and messaging pillars, the tone rules that decide whether you use questions, numbers or sentence case, the topics and terms this brand is trying to own, the product facts so a headline never promises something the page cannot deliver, and the pages you have already published so a new title does not compete with one of your own. The brain is permission-aware, so a client's keyword strategy stays inside that client's account.
Then it works end to end inside Sylvie: it generates a ranked set with the reasoning attached (the intent each option serves, where the term sits, the character count), pairs the winner with a matching meta description and social hook, and offers genuinely different angles when you want variants to test. You choose and edit in place, and the choices feed back into the brain, so the patterns your audience responds to compound instead of being rediscovered every quarter.
Each option is built around what someone typing that query actually wants, so the headline earns the click rather than simply containing the keyword.
Title tags come back at a length that survives the results page, so the part that does the persuading is not the part that gets truncated.
Options are generated from your tone rules and positioning, so the version that performs is also a version your brand would happily put its name on.
You get a set with the thinking attached: the angle, the intent, the placement. Choosing takes a minute, and the rationale survives into the client review.
It checks new titles against what you have already published, so a fresh post does not end up fighting one of your own ranking pages for the same term.
A quarter of headlines can come out of one pass without collapsing into a single formula, so the calendar stays varied and the brand does not read as a template.
A few of the jobs a team builds this agent for, every option written against your brand brain and the searches you are actually trying to win. The outcome is a headline that holds its ranking and its voice at the same time.
A page sits at position four with a title written three years ago and a click-through rate to match. The agent rewrites it for current intent and current positioning, so you pull more traffic out of a ranking you already earned without touching a word of the body copy.
An ecommerce client launches a category where every title is the same string with the SKU swapped. The agent writes a distinct, term-led title for each page, so the pages stop splitting relevance between themselves and each one competes on its own merit.
A quarter of topics is approved but every working title is a note to self. The agent turns them into publishable H1s and title tags in one pass, so writers start from a headline that already carries the angle and the term instead of inventing one at the end.
A new product page needs a line that carries the positioning, hits the category term and reads well in the results. The agent returns a ranked set with the trade-offs stated plainly, so the decision happens in one meeting rather than across three Slack threads.
A paid landing page test is scheduled but the proposed headlines differ only by adjective. The agent produces genuinely separate hypotheses (outcome, objection, proof, urgency) from the same positioning, so the test returns a finding you can act on.
A B2B brand wants traffic and refuses anything that reads like a listicle. The agent works strictly inside their tone rules, so the headline earns attention through specificity and nothing reaches the client that they would be embarrassed to publish.
A generator gives you twenty variations of the same guess, because it knows nothing about your brand or your search targets. This agent is built on your brand brain, so it writes from your positioning, your tone rules, your terms and your existing library, which is the difference between headline options and headline options you can actually use.
You build it. Sylvie does not ship a finished Headlines agent; this page is an example of what teams assemble on the platform. Sylvie helps you wire it to your brand brain and to the way your team briefs, reviews and publishes, so it fits your SEO process rather than replacing it.
It reads your positioning and messaging pillars, your tone rules, the terms you are trying to own, your current product facts and the pages you have already published, then writes every option against that picture. Because the brain is permission-aware, one client's keyword strategy never informs another client's headlines.
That is your call at build time. You can feed it the keyword targets your SEO team already produces, or wire it to the research step and let it start from a topic, since you decide what the agent takes as input and what it hands back.
The opposite is the point. Because it works from your positioning rather than from a list of proven formulas, it can produce a quarter of headlines that share a voice without sharing a structure, and it flags when an angle is one you have already used.
Yes. Every brand brain is isolated and permission-aware, so one client's positioning, keyword targets and published work never surface in another client's headline set. Each team member sees only the accounts they are cleared for, and nothing is pooled into a shared model.
Book a demo and we will map the workflows worth turning into agents, and show how each one runs on your brand brain.
Request a demo