Working with a reading: the plan, the issues, the chat

The overview opens with the SEO score and a verdict in words: Healthy from eighty, Good with room from fifty-five, Needs work below that. Beside it, three numbers: the actions this week, the time they take, and the pages read of the pages found.

The week is capped at five bundles on purpose. Everything else is one click away, but the Monday briefing is meant to be finished, not admired.

The action plan is three lists. This week holds the leaks worth your time. Next holds leaks too, just behind the five above; these are still costing traffic. Later holds housekeeping and template work, where nothing is costing traffic today.

Cards that share one fix surface are bundled into one sitting, and the line under a card counts edits rather than symptoms: so many edits, so many pages, roughly so many minutes. Each card names who can do it, You, Theme or Developer, and where: in the CMS on the pages themselves, in the template once for every page built from it, or on the server, the CDN or the code.

Every card has three answers, and each is recorded so the next reading knows what you meant.

• Mark fixed: the next reading confirms it, or a live check on the affected pages runs straight away and reports either gone from all pages checked or still open on some.
• Snooze for seven or thirty days: the card leaves the week and the counts, and comes back when the time is up.
• Ignore, with a reason: Not a problem here, False positive or Won’t fix. The card moves to Set aside, where the next reading keeps it rather than raising it again, and Bring back restores it.
• This is intended: on a card about noindex or a robots.txt block, one switch tells the next reading that the state is deliberate, and it stops being a finding.

Most fixes are done by someone else, so a card can leave the panel in the shape that person needs.

• Copy brief for the developer: the card as a ticket, with the pages and the change.
• Copy prompt for your agent: a prompt built from the catalogue and this reading, for an AI coding assistant. It is deterministic, costs no AI quota and works on the free plan.
• Export CSV and Export plan as Markdown, for the spreadsheet or the wiki.
• Draft AI fixes: titles, descriptions and alt texts written for the affected pages, within the plan’s monthly allowance.
• Redirect map and a ready-to-paste robots.txt block, where the reading found the pattern to close.
• Fix with Claude Code, on the Reports page from the Business plan: the audit as a SKILL.md, so a coding agent can apply the fixes in your repository.

The Pages screen is the raw table of the latest reading: status, title, description, H1, indexability, depth, issues, response time, word count, canonical, inlinks and outlinks, page type, and the Search Console columns when a property is connected. Filters narrow it by status, indexability, page type and issue, and the whole table exports as CSV.

Above the table, URL structure groups the site into templates, one row per pattern, with the share of pages, how many are in scope, noindex or canonicalised elsewhere, and what the site says about them. A page opens into its own detail: indexability with the canonical and any redirect chain, page data, the issues on that page, and the search queries it appears for.

Issues is every problem in the latest reading grouped by type, with a severity filter, a category filter and, for each type, what to do and the affected pages. Server-wide hygiene sits in its own group: one setting on the server each, showing on every page, not what the week opens with.

Compare with an earlier reading and every group shows what is new since, what is unchanged, what grew or shrank and what is gone. That comparison is what a second reading is for.

Changes lists what moved between your last two readings: issues fixed and new, pages added and removed, status, canonical, title, robots and sitemap changes, with a note where a fix came back or a redirect was removed. It needs a second reading and the Studio plan.

From Studio, a site can be read on its own every week, and from Business every day. Between readings, five alerts watch for a state that changed, never for a number that moved: the home page stopped being indexable, most pages became noindex, new server errors, robots.txt changed, and the score fell by ten points or more. Each needs a previous reading to compare against, so the first reading is always silent. Your own email switches decide what reaches you, and teammates keep theirs.

AI Chat answers from this site’s latest reading, not from general advice. Ask what to fix first and why, which pages are near-duplicates by template, why a page dropped, or for a developer brief on one card. The scope can be the whole site or one page. Each plan carries a monthly allowance of messages, shown in the sidebar meters.

Reports turns the latest reading into something to hand to a client: an Executive summary on every plan, a Full audit from Studio, Monthly progress and a Migration check from Business. A report downloads as a PDF from Studio and always has a shareable link; on Agency the PDF and the shared page carry no SilentStork branding.