Guides / Content team
Everything a content editor or marketer needs to build a library that personalises well. No code required.
Seven types cover most of what a B2B or SaaS site needs to say. Each type has a specific job.
Explaining one thing your product does. Not a list of features. One capability, written for one person.
Cost, plan breakdown, or value justification. Write a separate atom for each audience: one for developers, one for decision-makers.
One objection, one honest answer. Write faq atoms for the questions that come up on sales calls.
You vs. a named competitor or alternative approach. Honest comparisons convert better than vague claims.
Step-by-step instructions for a single task. One task per atom. Written in prose, not bullet points.
A specific outcome for a specific type of customer. Real examples work best. Plausible scenarios work early on.
A quote with context. The context (who said it, what role, what stage) is what makes it match the right visitor.
Seven rules that separate atoms that get matched and convert from atoms that sit in the library unused.
A pricing atom covers pricing. A feature atom covers one feature. Spectare assembles atoms together, so they do not need to cover everything individually. An atom that tries to say three things will rank lower in assembly than three atoms that each say one thing well.
Set the audience (developer, manager, executive) and buying stage (awareness, consideration, decision) before you open the editor. The framing changes completely. A developer at awareness needs to understand what the product does. An executive at decision needs to understand ROI and risk. Same product, completely different atom.
Spectare shows only a few blocks per page, three by default and up to six. If each atom is 600 words, the assembled experience is overwhelming. Write the essential version. You can always add a second atom that goes deeper if you need more coverage. Remember: your content appears on the page exactly as you write it. Every word the visitor reads in a content card is your word.
A stat makes an atom more credible and more specific. Setup time, price, a before/after number, a conversion result. Stats are displayed prominently above the prose. Even approximate numbers ("under 30 minutes", "from $19") anchor the copy.
Intent keywords are the search phrases your buyers use. "How does Spectare work", "website personalisation without A/B testing", "add dynamic content to WordPress". These feed the classification engine and help Spectare match atoms to the right visitors.
Bullet lists fragment the reading experience and make atoms feel like slide decks. Use full sentences and short paragraphs instead. Your content appears on the page word for word, not paraphrased, so every sentence needs to read naturally where the visitor sees it.
Write each capability as a standing fact. Avoid before-and-after or release-note framing ("we used to", "now also"): a visitor was not there for the before. And keep internal detail out, no API paths, function names, or code terms. Describe the value to the reader, not how it is built. The quality check flags both.
How your content reaches the visitor
Spectare uses two different approaches depending on which content component it places on the page.
Your words, copied exactly
Content cards and testimonials
The title and body of your atom are copied word for word onto the page. Spectare does not paraphrase, summarise, or rewrite. What you type in the editor is what the visitor reads.
Claude writes, using your atom as source
Headlines and calls to action
The headline and call to action are written fresh by Claude for each visitor, drawing on the atom you provided. The intent, claim, and evidence come from your atom. The framing is personalised for this specific visitor.
This is why atom quality matters. For verbatim components, weak copy stays weak on the page. For Claude-written components, a vague atom produces a vague headline. Write your atoms as if they will be read directly, because for the majority of placements, they will be.
Atoms move through five states from first draft to live on the site.
Draft
only editors and admins see it
In review
quality check results shown inline
Approved
staged, not yet live
Published
live in assembly immediately
An approver can also send an atom back to draft with a note instead of approving it.
The default state for new atoms. Only visible to editors and admins. Spectare runs an automatic quality check when you submit for review, flagging style issues, missing stats, and incomplete metadata before a human sees it.
Submitted for review. Approvers can see the quality check results inline. They can approve, reject back to draft with a note, or edit directly. Atoms in review are not live.
Approved but not yet published. Use this as a staging state: review in batches, then publish when ready. Atoms in approved state are not live.
Live and active. Spectare includes this atom in assembly for matching visitors. Publishing takes effect immediately. No deploy, no cache to clear.
Sent back to the author with feedback. The atom returns to draft and can be edited and resubmitted.
Automatic quality check
When an atom is submitted for review, Spectare runs a Claude quality check automatically on all plans. It checks formatting and length (bullet lists, word count), changelog or before-and-after framing, internal detail that should not be public such as code or function names, missing key stats on feature and pricing atoms, and incomplete metadata. The results appear inline on the atom list so approvers can see issues without opening the editor.
Images are selected and optimised independently from text atoms. Spectare learns which images convert best for each audience and stage combination.
Upload images from the admin under Images. Each image can be tagged with audience affinities (developer, manager, executive) and buying stage affinities (awareness, consideration, decision). These tags help Spectare decide which images are candidates for a given visitor before the selection algorithm runs.
Spectare uses a multi-armed bandit algorithm to select images. For a visitor with a known audience and stage, Spectare looks at the conversion rate history of each image for that combination.
90% of the time it serves the image with the highest conversion rate for that audience and stage. 10% of the time it serves a random alternative to gather data on images that have not been tested yet. Over time, the algorithm shifts weight toward images that convert and away from those that do not.
If a specific image must always appear with a specific atom (a product screenshot next to the feature it shows, for example), set the same Group ID on both the atom and the image. Locked pairs bypass the selection algorithm and always travel together. Everything without a Group ID is optimised freely.
When a visitor converts, Spectare records which image was shown alongside which atoms. This feeds back into the selection algorithm: images that appear at conversion get a higher score for that audience and stage, and are served more frequently to similar visitors in future.
Three tools under Atoms help you keep the library healthy as it grows.
Open Atoms, then Audit, for a read-only health check of the whole library. It shows your coverage across audiences and stages, which parts of the page are well supplied and which are underused, how atoms have performed over the last 30 days, and a learned-affinity view that suggests a re-tag when an atom converts differently to how you tagged it. Available on every plan.
Open Atoms, then Splits, and run the structure review to check your published atoms for the problem rule 01 warns about: an atom trying to make several claims at once. Accept a suggestion and the pieces arrive as drafts to review and publish, so nothing changes live without your say-so. Available on every plan.
When the audit surfaces a gap, or you already know the atom you need, Spectare can draft it from a short brief. The draft lands in the same review workflow as anything you write by hand.
Give your site’s URL at signup and Spectare drafts your starting library from your real pages instead of a generic template. On WordPress, Settings, Spectare Import repeats this any time. Imported atoms land as drafts in the same review workflow. See the import guide.
Library health
Start with 10 published atoms before you expect meaningful personalisation. Spectare needs enough variety to assemble different experiences for different visitor types. A new workspace starts either with atoms imported from your existing site, if you gave its URL at signup, or with ten AI-drafted starter atoms otherwise, so there is always a base to review and publish from.
Cover at least two audience types and two buying stages. A library of 20 feature atoms all written for developers at consideration will assemble the same experience for every visitor.
Check the analytics page regularly. The content gap detection agent runs nightly and surfaces intent patterns that are coming in from visitors but not matched by any current atom. These are your next writing briefs.
Retire underperforming atoms. Once an atom has at least ten impressions, if its conversion rate is running well below the average for its type, it is dragging down assembly quality. The quality agent flags the worst of these automatically and rewrites them as fresh drafts for your review.
Sign up for a free trial and Spectare generates your first 10 draft atoms automatically.