Zoviz technical SEO audit & programmatic page strategy

The JSON-LD Product schema claims an AggregateRating of 4.86 stars from 4,261 reviews. None of that rating is visible on the page. No star widget, no review count, no numerical average, and no reviews section matching the numbers. The same 4.86 / 4,261 figures are then repeated across dozens of service pages.

Why it matters

Google's structured-data guidelines explicitly prohibit marking up content not visible on the page. Repeating an identical rating across many pages amplifies the signal from "oversight" to "systematic spam." Consequences escalate from loss of rich results to site-wide manual actions under Spammy structured markup. For a branding platform whose entire SEO funnel relies on Product rich snippets, this is the single most exposed risk on the site.

Observed redirect chain for https://zoviz.com/about/:

Two compounding problems

(a) The canonical English URL redirects server-side based on a language cookie (TdwVuGzjzMvXrmPN=ar). Google's hreflang guidance is explicit: don't auto-redirect by perceived language; use a banner. When Googlebot crawls from a non-English IP or re-enters with a stale cookie, the English canonical bounces to a locale variant, diluting its ranking signals and flattening hreflang clusters. (b) The first hop drops to plain HTTP. HSTS saves most visitors silently, but the chain itself is 4 hops instead of 1, wasting crawl budget and tripping every SEO crawler's redirect-chain warning.

This is the issue most tightly correlated with the traffic decline shown earlier. /tr lost 86% of its traffic, /ar lost 53%, and both are homepages that cascade quality signals to their entire subdirectory. Manual inspection of those pages turns up a consistent pattern of half-localized content. The shell is translated, the substance isn't.

Taken together, this is the fingerprint of an automated translation pipeline injected into an English template without native-speaker review. Google's late-2025 and early-2026 core updates have both specifically targeted thin, programmatically-localized content, which is the most plausible explanation for the November 2025 peak and subsequent decline.

static-0.xml lists only the 192 English URLs as <loc> entries. The 12 locales per URL exist only as hreflang alternates inside those English entries. Google tolerates this; Bing, Yandex, Baidu, and DuckDuckGo heavily prefer explicit <loc> for every indexable URL. With 192 pages × 12 locales, roughly 2,300 indexable URLs are being under-declared to every non-Google crawler, and localized pages get far less crawl priority even from Google.

Separately, robots.txt disallows three URLs that are still listed in the sitemap:

• /hire-expert
• /website-creation
• /marketing-automation

Search Console flags these as "Submitted URL blocked by robots.txt", a self-inflicted crawl conflict.

Internal blog links are being written with a trailing ?srsltid query parameter. On a representative post, the link at XPath //*[@id="dynamic"]/div[2]/div[2]/div[1]/div[1]/div[3]/div[7]/p/a points to https://zoviz.com/blog/how-to-build-brand-awareness?srsltid=…, which Google will treat as a second, parameterized version of the canonical URL. The pattern is widespread inside the blog's internal-link graph.

Why it matters

srsltid is Google's Merchant Center click-tracking parameter; it's designed to appear on Google-hosted referral URLs, not on internal site links. When the site itself emits it, three things go wrong in sequence: (a) Google indexes both the clean and parameterized URLs, doubling the indexed footprint for every affected post; (b) ranking signals (links, engagement, dwell) split between the two URLs; (c) the resulting duplicate bloat sits precisely on the blog, the content surface most exposed to the Helpful Content algorithm's quality signals. At Zoviz's scale, this is not a cosmetic issue; it's quiet, cumulative index pollution.

Core Web Vitals risk: 101 of 120 images missing width/height

Before looking at specific attributes, the public field data makes the problem concrete. Google's Core Web Vitals Summary of this origin on phones as of 2026-04-11 (CrUX Vis) shows zoviz.com failing the CWV assessment at origin level on mobile. In other words, the CWV problem is not a one-page outlier. It is site-wide and already visible in the real-user data Google uses for ranking.

CrUX Vis, p75 distribution of user experiences for zoviz.com on mobile, 2025-06-22 to 2026-04-18. The vast majority of real-user sessions fall into the Poor band, a smaller share into Needs Improvement, and almost none into Good. The shape is stable across ten months, which rules out a one-off regression and points at a persistent, structural CWV problem.

Poor Needs Improvement Good

The homepage-level cause is easy to spot: 101 of 120 images have no width attribute and 110 of 120 have no height. Without intrinsic dimensions, the browser reserves no space before load, so content jumps as images lazy-load, directly inflating CLS and producing visible jitter during first paint. Combined with the 4-hop redirect chain in issue #02, LCP also drifts in the wrong direction. Fix: add width and height to every <img>, or set aspect-ratio in CSS. Trivial template change with a measurable CWV impact.

Heading hierarchy skips levels

The homepage jumps from <h2>Logo designs that match your industry</h2> directly to <h4>Real Estate logo</h4>. Breaks screen-reader navigation and dilutes topical structure. Change h4s to h3s.

Case-sensitive 404s

/About and /Pricing return hard 404s. Low-impact, but a soft 301 to lowercase catches typos and copy-paste errors from email and docs.