Creating content
Picking a channel, working through the five wizard steps, and what you can do with a piece once it exists.
Creating a piece moves through three screens: the channel picker, a five-step wizard for that channel, then the piece itself.
Step 1: Pick a channel#
The Create screen is a grid of channels, six of which are available:
| Channel | Best for |
|---|---|
| Blog | Long-form on your own site, and the main vehicle for closing topic gaps. |
| Medium | Long-form for an audience beyond your site. |
| Discussion, AMA and question formats. | |
| Quora | Answers to real questions. |
| Posts and carousels for professional context and B2B authority. | |
| X | Threads and single posts. |
Instagram and YouTube appear on the grid marked coming soon and cannot be selected.
Two of these are worth understanding beyond what the label suggests. Reddit is cited unusually often by AI systems, which makes genuine participation there more valuable than its audience size alone would imply, though it comes with strict community norms: promotional posting works against you and will frequently break a subreddit's rules outright. Quora earns citations out of all proportion to the effort involved, because question-and-answer content maps almost directly onto the way people phrase things to an assistant.
Content suggestions from the Content Scout also surface here, so you can start from a proposed idea rather than an empty channel. See Content suggestions.
Step 2: The wizard#
Choosing a channel opens a five-step wizard with a breadcrumb across the top tracking where you have got to.
1. Content Type establishes what kind of piece this is within the channel, since a Reddit discussion and a Reddit AMA are not remotely the same shape.
2. Select Topics asks which of your tracked topics the piece is aimed at, chosen from your Discovery Topics. This is the step that ties the piece to something you actually measure, which is what makes it possible to say later whether it worked.
3. Prompts narrows that to the specific prompts within those topics. Choosing prompts rather than a whole topic is the difference between "write about pricing" and "answer the three questions buyers actually ask about pricing", and the second brief produces a considerably better piece.
4. Craft is described on screen as choose keywords, then a title and a writing brief. Focus Keyword grounds the piece and is chosen from a Focus candidate list, with Citation Keywords and Citation Share showing what is already earning citations in that area, which is a better guide than search volume on its own. Keyword Strategy covers your secondary keywords, using Add another keyword and Add secondary keyword and press Enter. After that comes Craft a Title and then the writing brief.
The screen enforces that order deliberately, so you will see Choose a focus keyword above to craft a title, then Choose a title above to generate a brief, with Add a title and a brief to generate staying put until both exist.
Additional Context and Content Generation Criteria are where you add the things only you know, and this is the highest-value field in the entire wizard. A brief carrying your real numbers, your actual customer situations and your genuine constraints produces a draft worth publishing. A brief left generic produces one worth rewriting, and rewriting it will cost you more time than filling the field in would have.
5. Review & Generate is a final look before the draft is produced.
Some channels run a shorter wizard of Content Type followed by Craft, simply because there is less to decide.
The content wizard
- 1
Content Type
What kind of piece, within the channel
- 2
Select Topics
Which tracked topics it is aimed at
- 3
Prompts
The specific questions to answer
- 4
Craft
Keywords, then a title, then a writing brief
- 5
Review & Generate
A last look before the draft is produced
Step 3: The piece itself#
Once generated, a piece opens in its own view, reached from Content → Drafts or Content → My Content. This is a different screen from the wizard, and it is where the rest of the controls live.
Content quality rates the draft. Where a channel needs its own version of the piece you will see Adjusted for Medium, or Converted once that has been done, and if it fails, Could not prepare the Medium version is worth one retry before you assume the content itself is at fault.
Author and Custom author set the byline, and this is not cosmetic. Attribution is one of the things an engine weighs when deciding whether a page can be trusted, so a piece with no identifiable author starts at a disadvantage it does not need.
Copy all takes the text out, and the destination controls handle publishing. See Publishing.
Changing a piece after it is generated#
Two controls sit next to the quality score, and they differ in what tells the model what to do.
Improve this article works from the quality score. It appears when a piece scores below 70, and it fixes the specific rules the piece is failing. It costs nothing, and you can run it again for as long as the score stays below the threshold.
Revise with feedback works from what you type. Paste the feedback you were given, in your own words: name a section, quote a phrase you want changed, or paste a link you want added. Only what you mention changes. Everything else comes back exactly as it was, which is the point of using this rather than generating the piece again.
Feedback about the meta description or the URL slug will not be applied. Edit those fields directly.
A revision costs a fraction of what generating that piece cost, and the exact number is on the Revise button before you commit to it. You are not charged when the model concludes nothing needed changing.
Reviewing the changes#
Neither control edits your article on its own. Both show you what changed, one change at a time, with removed text struck through and new text in green. Accept or reject each one. Accept all takes everything still undecided, so if you have already rejected three of ten changes those three stay rejected and the remaining seven are accepted.
Anything you leave undecided counts as unchanged. Nothing is saved until you choose to save.
If a revision rewrote much more of the article than the feedback called for, you will see a warning above the changes. That usually means the feedback was broad enough to be read as licence for a rewrite. Reject the changes and try again with something more specific.
Version history#
Saving a reviewed revision creates a new version rather than overwriting what you had. The version number sits beside the date on every piece and opens the history, whether or not there is more than one version to see yet.
Revising with feedback, improving, and editing or replacing the image by hand each save a new version. Inserting an infographic or a carousel does not: those attach to the version you are on.
View opens a version read only, so you can read an older draft without changing anything, and copy it as Markdown, plain text or HTML. Compare shows what changed between any two versions, including an image that was replaced, and takes you to the first change.
Restoring an older version moves forward rather than back: it saves that version's text as a new version at the top of the list. You are asked to confirm first, and the dialog names the version it is about to create, because the number goes up rather than back. Nothing is ever deleted. The versions above the one you restored stay exactly as they are, so a restore can itself be undone by restoring again.
Best practices#
- Aimed at a measured gap, which is what steps two and three exist for.
- A brief containing what only you know. This is the whole difference between a draft that gets cited and one that reads like everybody else's.
- Answers one question thoroughly rather than surveying five of them adequately.
- Structured so a single claim can be lifted out without dragging the surrounding paragraph along with it, since that is exactly what an engine does.
- Sounds like your brand, which is what Brand Hub is for.
Next steps#
- Start from a gap rather than the channel grid. Gap analysis tells you what is worth writing.
- Read the pages currently being cited for that gap before you generate anything.
- Choose the channel the gap suggests. If Reddit threads are winning a topic, a blog post is addressing a different problem from the one you have.
- Pick specific prompts at Step 3 rather than accepting a whole topic.
- Write a real brief at Step 4, since everything downstream depends on it.
- Set the author.
- Check the claims before publishing. It goes out under your name, and an engine quoting a wrong number back at your prospects is a worse outcome than not being quoted.
What to keep in mind#
Seerly drafts and stores content. Publishing happens through WordPress or Webflow, or by exporting. See Publishing.
Draft housekeeping is deliberately conservative. An empty start, a draft with no title and no brief, is cleared after a couple of days. A draft you have actually worked on is kept for a year from its last activity before being treated as abandoned. And once a piece has been generated, the finished content itself is never deleted; only the wizard scaffolding behind it is eventually cleared.
Automated posting to community platforms is not offered. Those platforms restrict it, and participation that reads as automated damages the standing you were trying to build in the first place.
Related
- Content overviewHow content works in Seerly, from a gap in your data through to a published piece.
- Brand HubConfirm the brand names, voice and design language Seerly read from your site, so generated content sounds like you rather than like anyone.
- PublishingGetting finished content out of Seerly, through WordPress, Webflow, or export with posting instructions.
- Content suggestionsContent Seerly proposes based on your own gaps, and how your accept and reject decisions shape later suggestions.