Open Pages, choose a page, then choose Assessment → Assess this page. The free assessment reads the page’s public HTML and combines it with your saved audit, keyword map and connected Google data. It uses no AI credits or keyword lookups. The original HTML read is cached for ten minutes; the saved assessment records when its evidence was read.
Read the findings
Guided view starts with five cards. Open the full assessment for all ten sections. On a phone, one section opens at a time. Source labels distinguish measured HTML, dated website data, rules and your own answers. An item marked unverified is not a pass. Empty HTML does not prove that a page has no content: this reader does not run scripts.
The scorecard covers the existing fourteen On-page items without adding a new score. Reading level, keyword repetition and a possible services hub are writing or structure suggestions, not ranking targets. Schema guidance follows the page’s purpose; markup must describe real, visible content. Check the main action, phone readability and forms yourself, then save your answers.
The recommended title, description and H1 use the same fields as the Page view. Under Build on this assessment, choose a keyword and a writing step, then Review cost. Nothing starts until you confirm. A full assessment draft uses 6 AI credits and at most 1 keyword lookup: search strategy 1, rewrites 1, content gaps and outline 4. Summary polish is an optional extra credit. Your own AI key pays for writing without Arva SEO credits; keyword lookups still follow your plan. Google results are shared with Content Studio for 30 days. A double click reuses the same work. Failed writing returns its AI reservation; a completed search still uses its lookup if later writing fails. Choose suggested wording to put it in the shared editor, then use its Save button. Saving does not publish it.
Keep and share a report
Reports → Page assessments keeps saved assessments and earlier reads, even after old crawl rows are removed. A newer audit or changed saved recommendations marks the assessment out of date on the page list; the old evidence stays intact. This does not confirm a change to the live website.
Choose Create assessment report, select the sections and review any AI drafts before sharing. An unreviewed-draft warning does not block you. Tick the review stamp only after reading it, to include your name and date. The report freezes those findings, selected drafts and your saved recommended copy. Later changes do not rewrite it.
Work order and publication reports
Website audit → Set work order supports the recommended order or a custom order. Set priority labels and weights from 0 to 100 for priority, page importance and quick fixes; at least one weight must be above zero. Priority contributes Critical 100%, High 67%, Medium 33%, Low 0%. Importance uses its 0–100 score. Quick fixes contribute 100% minus estimated minutes divided by 240, capped at 240 minutes. Drag fixes or use Up and Down for a manual order; this takes precedence over weights. Clear manual order to use the weighted rule. Critical security findings stay first and critical findings keep their label. Severity and SEO Health never change. Existing assessments and reports retain their saved order; assess again for a new copy.
Activity → Publication webhook lets a person who can manage integrations create, replace or disable a signing key. A developer sends JSON with a unique event_id, a page url on this website and an optional description. Sign timestamp + "." + the exact JSON body with HMAC-SHA256 and send X-Arva-Timestamp and X-Arva-Signature. Signatures expire after five minutes; repeated event IDs return the same entry. Up to 100 new events per website per day. The key is shown once, stored encrypted and never belongs in public JavaScript. Replacing or disabling it invalidates the old key.
These entries are labelled reported publication, not verified changes. They may offer a 90-day check and exclude the page from unchanged controls; nothing starts automatically. They do not publish content, verify fixes or use paid data. The app menu previously named On-page SEO is now Pages; the audit’s On-page SEO category and existing links keep working.
- Copy formatted text copies real tables for Word or Google Docs. If your browser only permits plain text, the confirmation says so.
- Download DOCX creates an editable Word document from this report’s frozen sections. It uses no PDF allowance or AI credits.
- Delivery options → Google Docs creates one private editable document in your own connected Google account. Connect Google for document creation if needed; only access to files created or opened through Arva SEO is requested. No sharing permission or email is added. Reopening returns the same document. If Google does not confirm an upload, check your Drive; Arva SEO does not automatically create another copy. Google may adjust tables. Changes in Docs do not change the saved report.
- Delivery options → Public share link creates a link readable by anyone who has it for 1 to 30 days. Check the frozen report before sharing. The link is shown only after creation, is marked noindex for search engines, and can be revoked. Creating another replaces the old link. Downloaded or copied content cannot be recalled.
- Assessment templates save the selected sections, their order and a plain-text introduction for this workspace. Use Up and Down to reorder sections. Using an existing template name replaces its settings; deleting a template does not change reports already created. Current plan branding still applies. Reports, copies, DOCX, PDF, portal and public views use the same frozen introduction and section order.
- If a delivery step fails: when Arva SEO could not reach Google at all, nothing was created and you can simply try again. If Google did not confirm an upload and you checked Drive and found no document, choose I checked Drive, allow another try. Creating a new public link while one is live asks you to confirm first, because the old link stops working for everyone who has it. A signed publication webhook accepts up to 100 new events per website per day; further signed events that day return 429 Too Many Requests.
SEO File → Page assessments lists the latest saved assessment for each page. It includes the read date, keyword and whether it was mapped or suggested when assessed, first priority, top findings, saved recommendations, manual-review answers, draft-writing status and a link to the saved assessment. A newer audit or changed saved recommendations marks it Out of date; No known change does not mean the live page was rechecked. Summaries are shortened; open the assessment for full evidence and history. Opening the sheet uses no AI credits, keyword lookups or new website tests. Excel and Google Sheets exports include the same sheet under your existing SEO File export permissions. Exports are dated copies; they do not update after export. Private writing drafts and manual answers stay within your authorized workspace or the export you choose to create.
- Generate PDF uses one report from the existing monthly PDF allowance. Downloading the same PDF again is free. A failed generation returns the allowance. Branding follows your current report settings and plan.
- Share in client portal makes only that report available to the website’s assigned client. It sends no email. Remove from client portal stops portal access. Assigning the website to a different client does not share older assessments with that client.
A web report is free. The PDF includes a cover, selected sections, sources, how the page was checked and what could not be verified. Review the evidence before sharing it.
What still needs checking
Link suggestions use exact phrases found in retained text and the existing keyword map and crawl. When fewer than three can be supported, the assessment says what is missing. Pages competing reuses Search Console, rank swaps and mapped-keyword evidence; missing access is not a clean bill of health. Competitor gaps need readable Google results and a separate writing step. Drafts need human review. Use matching Search Console and Analytics periods to judge progress; ranking or enquiry gains are never guaranteed.
Observed changes and content depth
Activity includes observed plugin/theme version changes, WordPress version changes and reported speed-fix switches. Trying affects administrators only; a request is not a confirmed visitor change. First inventory, missing versions and missing crawl evidence do not prove an update. Dates are observations, not publication dates or proof of ranking effects.
One word-count review rule applies to new audits, On-page checks and assessments: 300 on home and service pages, 600 on posts and 200 elsewhere. Service pages are identified by Service markup or a service/practice-area address. Length is only a review signal; add useful detail, not padding. SEO Health v10 applies to new audits after release; older scores stay as saved.
The 90-day check
After a change from an assessment is marked done, open its Assessment and choose Start the 90-day check. Confirm the selected pages have been published. Both Search Console and Analytics need a complete 28-day saved baseline ending before the first recorded change. Missing or stale data is not zero. Up to 20 changed pages and 10 similar unchanged pages are kept with their selection reason; unlogged changes still need your review.
Every plan, including Free, gets day 30, 60 and 90. The baseline is kept independently of rolling history. Checkpoints read dated 28-day windows and appear in your notifications and Activity log. Routine emails follow your preferences; they do not bypass the essential-only default. Paid plans can save private checkpoint reports. PDF generation uses the normal report allowance when requested; nothing is shared automatically. Stop check stops future checkpoints.
Compare changed pages with unchanged pages. Controls changed after the baseline are excluded. Read Organic Search only and separate brand from non-brand reported queries. Last-year page comparisons need sufficient plan history and saved data. Google updates appear as light chart bands, separate from your work, with the list’s last-checked date. An empty list does not confirm there was no update. Record paid ads, promotions, seasonal events, site moves and other marketing in the Activity log; a large Paid Search change prompts you to add a note. Review links with recorded first-seen dates and other site changes. Seasonal peaks and Google updates can overlap the window; none of these observations proves cause. Typical timing ranges are editable by staff, never promises.
Assess again reads the free evidence afresh and keeps written drafts with their original dates. The comparison separates verified audit fixes, new and unchanged findings, missing findings that are still unverified, passing copy checks and whether live copy matches saved recommendations. Task verification feeds the done marks. Old snapshots without saved health or full findings remain explicit about that missing evidence. Review retained drafts and manual answers after a page changes.
Completed website crawls also compare matching fetched addresses with the preceding completed crawl. The Activity log groups observed changes to titles, descriptions, main headings, canonical addresses, indexing instructions, structured data types, HTML size and HTTP status. These dates are observations, not publication dates or proof of who changed a page. Missing or incomplete evidence is not a change. Comparison pages with observed changes are excluded from 90-day controls; if a saved group is incomplete, controls remain unverified.