What it reads, and what it does not
The tool fetches the single page you give it — the homepage is the usual choice — and reads three things from it: the title, the meta description, and the body text. From those it works out what the business sells, roughly who it sells to, and the words it uses about itself. The ideas come from that and nothing else.
It does not crawl your site, follow links, read your sitemap or look at anything behind a login. It does not search the web, read the results page, or look at what your competitors publish. And it does not know a single search volume, which is the most important thing to understand about it: there is no keyword data behind these ideas, only an informed reading of your own page.
That sounds like a limitation and it is, but it is also what makes the output usable. A topic list generated from keyword volumes tends to be a list of phrases rather than articles: technically searched, impossible to write with any conviction. A list generated from your own page is a list of things your business has standing to say.
- Reads: the page's title, meta description and visible text, up to about four thousand characters.
- Ignores: your sitemap, your other pages, your competitors, and anything behind a login.
- Never claims: search volume, keyword difficulty, traffic estimates or a ranking prediction.
What to do with the five ideas
Read the keyword before the title. The title is a suggestion and you should rewrite it in your own words; the keyword is the claim about what someone is actually looking for, and it is the part worth agreeing or disagreeing with. If a keyword feels wrong for your business, the idea is wrong, however good the title sounds.
Then read the intent. An informational idea wants explaining; a commercial one wants comparing. Those are different articles with different shapes, and writing one as the other is the most common way a well-chosen topic turns into a page nobody finishes. A page that answers "how does X work" with a pitch loses the reader in the first screen.
Keep two of the five. Five ideas from one page will usually contain one you have already covered, one that is slightly off, and three that are genuinely useful — and of those three, the two you can write with a specific opinion are the ones to start with. Run the tool again on a service page rather than your homepage if you want ideas about one part of what you do.
What separates a topic from a keyword
A keyword is a signal that someone is looking. A topic is a decision about what to tell them. The gap between the two is where most content programmes fail: the keyword research is fine, the calendar is full, and every article is a long, hedged answer to a question nobody asked in those words.
A good topic has three properties. It names a reader — not "businesses" but the person who has this problem on a Tuesday. It promises something specific that will be clearer or more possible by the end. And it is something you can say with conviction, because you have done the work and can name what goes wrong.
The last of those is the one a tool cannot supply. A generator can tell you that people search for a comparison between two things you sell; only you know which of the two you would actually recommend to whom, and that opinion is the article. Everything else is context around it.
- Name the reader, not the market.
- Promise one thing that will be clearer by the end.
- Say the part you would only say out loud — that is the article.
The same step, inside PrismDraft
This tool is the first step of the product, with the rest cut off. In PrismDraft it keeps going.
- 01
The project remembers
Your website becomes a project with a brand voice and an image style drafted from your own pages, so the next twenty ideas are shaped by it rather than by one page read once.
- 02
Ideas wait on a board
Each suggestion keeps its keyword, intent and topic cluster in Planning until you pick one, and a keyword the project already covers is never suggested twice.
- 03
The outline comes next
A heading per section with the job it does, its keywords and a target length. Free to generate, free to change, and approved before any prose.
- 04
Then the article
Written section by section against the outline, with an image for each section and alt text, for about 7 credits on the defaults — quoted before the run starts.
- 05
And a person says yes
It waits in review. Only an approved article can be published, to WordPress, Ghost, Webflow or Shopify.

Questions
Is it really free?
Yes, with no account and no card. There is a limit of five calls an hour from one address, because each one costs us a model call that nobody is paying for.
Do you store my website address?
No. The page is fetched, read, and the ideas are returned. Nothing about the request is kept, which also means you cannot come back for the same answer later — copy what you want to keep.
Why are there no search volumes?
Because we have no source for them. A number invented to look authoritative is worse than no number, and search volume from an unnamed source is exactly that. Bring volumes from a tool that actually measures them.
Can I use it on a competitor's site?
It will work on any public page, but the ideas will be about their business, not yours. Reading your own service pages one at a time is a better use of it.
Why only five ideas?
Because five is about as many as a single page can honestly support, and because a list of thirty is a list nobody reads. PrismDraft itself produces as many as a publishing plan needs, from a project rather than from one page.
What if my site has very little text?
The tool refuses rather than guessing: a page with almost no readable text answers with an error. Try a service or about page instead of a landing page that is mostly images.