Optimize

To do

One ranked list of what to do next.

Fact check findings, alerts, page fixes, citation gaps, Search Console opportunities, earned sources and prompt suggestions each have their own screen. To do gathers the open ones into one list, highest priority first. It keeps no status of its own: every button changes the record on its own page, so nothing is tracked twice.

• A priority you can read: scores from different sources measure different things, so they are never added up. Each item gets a High, Medium or Low band from a stated rule for its kind, and its reason line states the input behind it.
• Act from the list: mark a page fix applied, accept or dismiss a citation gap or a prompt suggestion, dismiss a Search Console opportunity, or acknowledge an alert, without leaving the page. Fact check findings and earned sources open on their own pages.
• Counted once: a citation gap is hidden while a page fix built from it is on the list, so the same work does not appear twice.
• Split by where the work happens: tabs for your own site, other sites and tracking, a New mark on anything created in the last 7 days, and a count of open High-priority items on the To do tab.

Every project member can read the list; owners and editors see the buttons. To do has no plan gate of its own: it lists the open records the project's sources already hold.

Headline Agent

On-Page Optimization

Recommendations are only useful if they’re specific, and safe to ship. The On-Page Optimization agent reads your live page alongside the competitor pages AI engines actually cite for your prompts, then hands back a ranked worklist of concrete edits: which page to fix first, the proposed change, and drafts you can review for your CMS. Each retained change has a quote matched to saved source text. After you mark a change applied, follow-up observations compare visibility with the 30 days before it.

• A ranked worklist: scored by impact, effort, and confidence, so you always know the single highest-value page to fix next.
• A page-level work queue: prioritize known owned pages by actionable issues, review linked evidence, inspect saved diagnostic findings and follow draft or applied status. The saved inventory is bounded and clearly labels its coverage; it is not a complete crawl of your site. Explore the page issue queue.
• Drafts for implementation: approve reviewed drafts for handoff in batches of up to 25, copy proposed text, or export a Markdown or PDF task packet with all saved draft formats, source quotes and a review checklist. Task exports use the project owner's data-export access, included on Starter, Growth and Pro and during the product trial.
• Evidence for review: saved source quotes support each retained change. The draft itself still needs checking, with additional human-review flags for regulated topics.
• Observe subsequent changes: mark a change applied after implementation, then compare later visibility with its baseline. These observations do not establish that the edit caused a change.

WordPress

A new draft in WordPress, never a publish.

Send a finished article into your WordPress site as a new draft post, then read it and publish it there yourself. Each send creates a new draft. DiscoveredBy never publishes, edits or deletes a post in WordPress, and never touches anything already there.

• Connect once, as a Contributor: the project owner connects a site on the project's own domain with a WordPress Application Password. We recommend a Contributor account, which can create drafts but cannot publish, so WordPress itself enforces the draft-only rule. The password is stored encrypted and never shown again.
• What arrives in the draft: the article's title and content, its meta description as the excerpt, and the suggested slug when it passes our check. Images are not sent; each one becomes its alt text. Page fixes from Optimizations stay task packets and are never sent to WordPress.
• No accidental duplicates: sending the same version again is refused while an earlier send created a draft or may have. A send that fails after WordPress may have received it reads Outcome unknown, so you check WordPress first; Send another copy makes a second draft on purpose.

Connecting a site and sending drafts follow the project owner's plan: Starter, Growth and Pro.

Site audit

See your site the way a crawler without JavaScript does.

A site audit crawls your own site as DiscoveredByBot, from its robots.txt, its sitemaps and the internal links on each page, and checks up to 1,000 pages in one run. Crawlers that do not run JavaScript only ever see the raw HTML a page delivers, so each page is also loaded once in a browser and the two are compared.

• One set of checks per page: each page it checks gets declared crawler policy, delivered HTML, structure and metadata, response size and timing, and the browser render. Each check reads Check passed, Needs review or Unknown, with its saved evidence.
• What only scripts add: a page whose text or same-host links grow substantially only after its scripts run, or whose title, main heading or robots directives change when they do, is flagged for review.
• Straight into your work queue: your newest finished audit adds one item per inspected page to the page issue queue, and a page checked in an earlier audit is compared with its previous result.
• Read-only and honest about it: requests to common third-party analytics and tag-manager services are blocked during the render, so an audit does not add page views to those tools, and DiscoveredBy does not edit your site. The structure, response and render checks are heuristics from one lab request or browser load, not Core Web Vitals, a citation prediction or a grade.

Starting an audit is for owners and editors on Starter, Growth and Pro; every project member can read saved audits.

llms.txt Advisor

A compliant llms.txt, written out for you.

llms.txt is a short Markdown file at the root of your domain that gives AI engines a curated map of your site. The llms.txt Advisor fetches your live file, checks it against the llmstxt.org spec, then has a model draft a compliant replacement for you to publish.

• Checked by rules, not a model: the checklist is plain code, so the same file always gets the same result. Only an empty or missing file, or one that does not open with a title line, fails; everything else can at most raise a warning.
• A file ready to copy: the proposed file appears with a Copy button. Nothing is written to your site; putting it live is a step you take yourself.
• On demand or weekly: owners and editors can run a check on demand, and the project owner can turn on a weekly automatic check.
• A structural check: it checks the file's shape against the spec. It does not follow the links to confirm the pages still exist, or judge whether a summary or a link description is any good.

Available on Starter, Growth and Pro, following the project owner's plan.