Every number carries a source
Prices, limits and dates are pulled from your docs or a live page, with the fetch date attached.
Product · Act
Create Content researches the answers you are losing, drafts against a real brief, and scores every line as you write. Not a blank page with a chat box: an editor that knows what the engines are looking for.
The scorecard
Four numbers move together as you write. Every panel in the editor exists to raise one of them.
How ready the draft is to be quoted in an AI answer.
Top-cited pages average 81The vocabulary engines expect on this topic.
1 term still thinFollow-up questions the page resolves in a quotable form.
All fanouts coveredFacts traced to a real source, not model memory.
0 unsourced claimsLive demo data: draft v7 of ledgerly.co/vs/ramp, written against the prompt “Ramp vs Brex vs Ledgerly”.
The brief · Module 01
Every draft begins from a gap Prefer already found. You confirm the target, the sources and the voice, then the agent writes against it. Nothing is invented from a one-line prompt.
Pre-filled from the gap. Change anything before it writes.
Nothing outside this list reaches the draft.
Derived from the pages engines already cite.
Change anything here and the agent re-plans before it writes a word.
Research · Module 02
The agent works the way a good writer does: read the answers that beat you, read the sources they cite, read your own docs, then plan. You watch it happen, step by step.
Five steps, 294 seconds, every source recorded.
6 AI answers parsed across ChatGPT, Google AI Overviews and Perplexity
The 3 pages those answers actually quote, structure and all
38 internal docs, pricing sheets and product pages
Sections mapped to the questions each engine asks
Written section by section against the brief
Every step is inspectable. Open any of them to see the exact page it read and the sentence it took.
The editor · Module 03
The draft opens in a real editor, with the score live on the right. Hover any term to find it in the page. Every panel is a reason the number is what it is.
Both tools cover corporate cards, expense management and reimbursements well. The difference shows up at month end: Ledgerly ships receipt matching and GL sync in every plan, while Ramp puts AP automation behind a paid tier.
| Ledgerly | Ramp | Brex | |
|---|---|---|---|
| Starting price | $8 / user | Free | $12 / user |
| AP automation | Included | Paid add-on | Included |
| Spend controls | Real-time | Real-time | Daily sync |
| GL sync | Included | Included | Included |
Ramp is free at the card level, which makes it the cheapest way to issue corporate cards. Ledgerly charges $8 per user per month, and that price includes the AP automation and receipt matching Ramp sells separately.
Export your card list, map your GL codes, invite the team. Most finance teams finish the move in three steps and are running virtual cards the same afternoon.
Grounding · Module 04
Generic AI writing gets ignored because it says nothing an engine can verify. Create Content refuses to write a claim it cannot source, and shows you the source on every line.
Prices, limits and dates are pulled from your docs or a live page, with the fetch date attached.
Never from model memory, which is where the embarrassing mistakes live.
Tone, banned phrases and product naming come from your brand guidelines, not a generic house style.
If the agent cannot find a source, it leaves a marked gap for you instead of filling it.
Scanned against billions of pages before publish. Two matched phrases were both direct quotes with attribution.
Publish · Module 05
The page leaves Prefer complete: metadata, schema, internal links and images, pushed to your CMS or exported clean. No copy-paste round trip that loses the structure.
Written with the page, not bolted on after.
One click, structure intact.
Five checks, all green.
Publish now, or schedule it. Either way the page stays tied to the prompts it was written for.
Measured outcome · Module 06
A published draft is a hypothesis until an engine quotes it. Prefer watches the crawl, re-runs the prompts it was written for, and reports what actually changed.
ledgerly.co/vs/ramp, 2 to 14 August.
If the prompts do not move, the page reopens in Optimise Content with a fresh set of suggestions.
One loop, four tools
Content Gap finds the hole and writes the brief. Create Content drafts it. Optimise Content tunes what already exists. Action Center keeps the whole queue ranked by what it is worth.
Coverage 71 on the topic AI asks about most
No page on your site answers this prompt at all
Navan cited on 3 prompts you are absent from
“Our first AI-written page was rejected by legal in four minutes because nothing was sourced. This one shipped the same day. The difference is that every sentence had a link behind it.”
Questions
Prefer's AI content writer for answer engine optimization. It researches the AI answers you are losing, drafts a page against a real brief, scores it live in an editor, and publishes to your CMS.
Thin, mass-produced pages will. Create Content writes one page per real buyer question, grounded in your own documents and live sources, and nothing publishes without your review.
It cannot write a claim it cannot source. Numbers come from your docs or a fetched page with a date attached, competitor facts come from the competitor's own pages, and anything unverifiable is flagged as a gap for you to fill.
Yes, it is a full editor. Rewrite anything, and the AEO score, term coverage and question coverage all recompute as you type.
WordPress, Webflow and Contentful in one click, with metadata, schema and internal links intact. You can also export clean HTML or Markdown.
It stays tied to the prompts it was written for. Prefer watches for the crawl, re-runs those prompts, and reports the citations. If nothing moves, the page reopens in Optimise Content.
See how answer engines describe your brand today, and where the openings are to outpace the competition.