Freshness and Web Vitals

Two page-level factors come up again and again when a competitor’s page is cited instead of yours: it was updated more recently, or it loads and responds faster. AmICited measures both, for your domain and for your tracked competitors, and places them next to the positions AI engines cite pages at. Two pages under Audit cover them: Freshness and Web Vitals. Each prompt also gets two sensitivity scores that tell you whether freshness or speed is likely to matter for that question at all.

Freshness#

Audit → Freshness is built from a daily crawl of your sitemaps and your competitors’ sitemaps. It shows how many URLs each host adds, removes and marks as updated, so you can see who is out-publishing whom, and where.

• Sitemap churn. One bar per day, segmented by host: added above the axis, removed below. Click a segment to see the URLs. Switch between Added, Removed and Net.
• Freshness leaderboard. A single 0 to 100 freshness index per host, built from four sub-scores (publish rate, refresh rate, recency and stale share) and size-adjusted, so a small site that ships steadily is not beaten by a large one that does not.
• URLs by last update. How many URLs each host last changed within 7, 30, 90, 180 and 360 days, or longer, from sitemap lastmod dates. “No date” means the sitemap states no lastmod: unmeasured, not old.
• Directory comparison. URL counts per path prefix, as a treemap you can drill into and color by freshness (the share of dated pages updated in the last 30 days).
• URL changes. Every URL added to a sitemap, dropped from one or marked as updated, with CSV export.
• Freshness vs citation position. Every URL cited for your domain’s prompts (not only competitors’) plotted by page age against its best citation rank. Page age comes from a date the page states (structured data, page metadata, page markup, publication date or HTTP header); a date taken only from the sitemap is marked as a weak signal.

History starts the day tracking begins for a host, so the 90-day view fills in over time. Re-check crawls a host’s sitemaps again on demand (0.005 credits), at most once a day. Per-competitor page counts and 30-day change are also on Competitors → Competitor sitemaps.

Web Vitals#

Audit → Web Vitals uses Chrome UX Report (CrUX) field data: how real Chrome users experienced the page over the last 28 days, at the 75th percentile. It covers LCP (main content load), INP (responsiveness), CLS (layout stability), FCP (first content) and TTFB (server response).

• Competitor comparison. Your homepage’s score and Core Web Vitals status against your tracked competitors’ homepages.
• Your pages by Web Vitals. Every page of your domain that AI engines cite, sorted by how often it appears in prompt sources, so you fix the pages that matter first.
• Performance impact. For each metric, your cited pages spread by speed against the position engines cite them at, filterable by engine, with a trend line. A flat line means the metric has no real bearing on your positions; a steep one means it does.
• What moves your position most. Which metric most tracks your citation position. Treat it as a hint, not proof: top-10 data is noisy.

Refresh Web Vitals queues a new import; URLs are re-checked gradually and it can take several hours. CrUX only has data for pages with enough Chrome traffic, so low-traffic pages may show the origin’s numbers or nothing at all.

Prompt sensitivity scores#

On the Prompts page, the Sensitivity columns show, for each prompt, how much its Google top 10 leans on fresh pages and on good Core Web Vitals. Both are computed from one observed snapshot of the Google results for the prompt text. They describe the result set, not why pages rank.

• “Good” uses the standard bands: LCP ≤ 2.5 s, INP ≤ 200 ms, CLS ≤ 0.1, FCP ≤ 1.8 s, TTFB ≤ 800 ms.
• A score needs at least 5 usable pages, and at least 60% of the top 10 must be dated (or have CrUX data). Otherwise the cell explains that coverage is too low to score.

Worked example#

For “best project management software”, 9 of the Google top 10 have a usable date and 7 of those were updated in the last 30 days: 7 ÷ 9 × 100 = 78. That is a freshness-hungry prompt, and a page last touched a year ago will struggle. A prompt scoring 10 is answered by evergreen pages; refreshing yours will not move much there.

How to act on it#

• Refresh where freshness sensitivity is high. Update the page that should own the prompt with real changes (new data, current year’s pricing, new sections), keep lastmod and on-page dates truthful, and log it as an annotation.
• Fix speed on cited pages first. Start at the top of “Your pages by Web Vitals”. Slow TTFB and content that only appears after heavy JavaScript hurt AI fetchers most, because retrieval bots fetch under tight timeouts.
• Watch competitors’ churn. A competitor suddenly adding hundreds of URLs in one directory is usually a content push aimed at the same prompts you track.

Related#