Audit: Content and Speed
Two categories that often turn out to be the real reason for weak positions: the content of the page and its loading speed. In the report, these are the categories "Content" and "Performance". In addition to checks of the page code, this includes Google PageSpeed Insights data on speed and checking whether the site is cited by AI search.
Content
| Finding | What it means | What to do |
|---|---|---|
| Little content (less than 150 words) | The page has nothing to respond to the query | Expand the topic or close the page from indexing if it is a service page |
| Content duplicate | The text matches another page on the site | Merge, make unique, or specify a canonical address |
| Keyword stuffing | The phrase is repeated unnaturally often | Reduce it — see keyword density |
Thin pages are the most common mass finding among online stores: cards with one sentence of description. This is not a reason to write a thousand characters for each, but a reason to decide whether all of them are needed in the index.
Speed
- Slow response (more than three seconds) — a server problem, not a layout issue: optimizing images won't help here.
- Heavy page (more than 2 MB HTML) — usually a consequence of embedded data or a giant menu.
- Response without compression — enabled by one server setting and immediately reduces traffic significantly.
- Blocking scripts in head — delay rendering: move them down or mark them as deferred.
- No meta viewport — the page does not adapt to mobile, and search evaluates the mobile version.
- Images without width/height — layout jumps during loading, see Core Web Vitals.
Core Web Vitals and Mobile Speed
These three checks take data from Google PageSpeed Insights. The audit measures the main page and several pages from different sections (three in the free check, six in the full audit), all in the mobile version. The detailed block "Speed and Core Web Vitals" with numbers for each page is described in the article how to read the audit report.
| Finding | Level | What it means |
|---|---|---|
| Core Web Vitals not passed | Error | According to real Chrome visitors' data over 28 days, at least one metric is in the red zone: LCP (loading of the main block, norm ≤ 2.5 s), INP (response to clicks, norm ≤ 200 ms) or CLS (layout shifts, norm ≤ 0.1) |
| Core Web Vitals on the edge | Warning | At least one metric is in the "needs improvement" zone — slightly below the green zone of Google |
| Slow loading on mobile | Warning | Lab measurement on mobile: PageSpeed score below 50 out of 100 or the main block appears longer than 4 seconds |
In detail, the problems indicate the metrics that did not meet the norm. If there is little data on the page itself and Google provides statistics for the entire site, the Core Web Vitals issue is reported once, marked as "for the entire site". If there are too few real visits from Chrome, the Core Web Vitals checks will not be performed at all, and only a lab measurement will remain.
What to do: start with the items in the "What will speed up the most" block in the report — there are fixes with the greatest time or weight gain. For LCP, compressed images in WebP or AVIF and a fast server with caching usually help, for INP — less heavy JavaScript, for CLS — reserved space for images and banners in advance.
Content and AI Search
The "Content" category also includes the results of visibility checks in AI search: the audit asks Perplexity questions about the site's topic and a question about the brand.
- AI search does not cite the site (warning) — in none of the answers to questions on your topic did the site make it into the sources. The details list the sites that AI cites instead of you. Pages with direct answers to customer questions (FAQ, comparisons, prices, instructions), facts and figures, microdata, open access for AI bots, and mentions of the site on authoritative platforms help.
- AI does not know your brand (note) — when directly asked about the company, AI did not find or mention your site. Mentions of the brand on third-party sites and an "About Us" page with facts are needed: who you are, where, since what year, what distinguishes you.
- No llms.txt (note) — see indexing issues, the section on access for AI search engines.
How to Distinguish a Duplicate from a Similar Page
The audit marks pages with nearly identical text as duplicates. But there is a trickier case: pages differ in text and are the same in meaning — cards of the same product in different colors, cities in a service template. Formally, there is no duplicate, but they compete just the same. This is resolved not by the audit, but by cannibalization check and grouping of queries.
Where to Look Next
Detailed analyses: page load speed, text coherence, article optimization for the query. If the pages are technically fine but there is no traffic — diagnostic checklist.