Phương pháp

What this site is. SEO2AIO is a research and information tool: it measures publicly accessible facts about a website and turns them into a graded report with a short, prioritized fix list — never professional advice and never a guarantee of rankings or traffic.

The crawl. When a URL is submitted, our crawler fetches the homepage (timing the server response), robots.txt, the XML sitemap, /llms.txt, the http:// and www/apex host variants, one deliberately non-existent URL (to test 404 handling), and the domain's public RDAP registration record. Fetches run from our own server with a stated user agent (SEO2AIOBot) and a hard byte cap; nothing else is accessed. Each review stores its crawl timestamp, and figures always refer to that crawl — sites change.

The checks. 45+ deterministic checks across ten categories: search visibility, titles/content, technical SEO, speed, mobile, structured data, AI & answer engines, local search, authority & trust, and accessibility. Every check records pass / partial / fail with the evidence it saw (the actual title text, the measured milliseconds, the exact robots.txt rule). Checks that cannot apply — local-search signals on a non-local site, image alt text on a page with no images — are excluded from scoring rather than guessed.

Scoring: two takeaway scores. Checks carry point weights inside their category and category scores are 0–100. Each category then feeds one or both of the two takeaway scores with pillar-specific weights: SEO (classic search visibility — search basics 18%, titles/content 16%, technical 12%, speed 12%, structured data 10%, mobile 9%, local 8%, authority 8%, accessibility 7%) and AIO (AI & answer-engine visibility — AI/answer-engine checks 30%, structured data 15%, titles/content 15%, search basics 12%, technical 8%, accessibility 8%, speed 6%, mobile 6%). Unmeasured categories are excluded and the remaining weights renormalized — never guessed. Grades per score: A ≥ 90, B ≥ 80, C ≥ 70, D ≥ 60, F below. Zones: green ≥ 80, yellow 60–79, red < 60. The "strong-ranking profile" compares each score with 85 — the score a site meeting the practices that consistently strong-ranking sites follow would earn on these same checks. It is a measured yardstick, not live competitor SERP data.

Recommendations and verification. A full review issues at most six prioritized recommendations, chosen deterministically from the highest-impact failing checks (plus clearly separated optional service pointers). Resubmitting the site re-runs the checks and verifies each recommendation — completed when the change is detected, "not detected" when it isn't, "needs time" when a crawl alone cannot confirm it — and your scores update from the fresh measurements. Recommendation sets are refreshed periodically rather than reinvented on every submission, so the plan stays stable while changes take effect.

Traffic Strength and Overall Digital Authority. Alongside SEO and AIO, each review shows Traffic Strength — a log-scaled 50–100 score computed directly from the domain's Tranco global popularity rank (#1 → 100, #1,000 → 76, #100,000 → 58, #1,000,000 → 50; outside the top million it reads “below the measurable range”, never an invented number) — and Overall Digital Authority, the composite of every measured category including popularity, domain age and trust signals.

Prominence. Real-world prominence is the strongest visibility signal there is, so the authority category includes each domain's global popularity rank from the public Tranco research list (top 1M domains, aggregated from major traffic datasets). This check only ever adds points — not being listed never subtracts, so small and local sites are not penalized. For domains in the global top 1,000, discovery-aid checks (sitemaps, schema markup, answer formatting and similar) are shown as informational rather than scored: at that level of prominence their absence demonstrably does not limit visibility, and the review says so on each affected row. Genuine defects — broken HTTPS, noindex, soft 404s, mixed content, accessibility — always count for every site.

What we do NOT measure. Backlink counts and third-party authority scores (commercial link-index data), live rankings for specific keywords, Google Business Profile data, field Core Web Vitals, and anything behind a login. Where a category depends on such data, the review says so explicitly instead of inventing a number.

AI assistance. The wording of initial-review summaries and recommendations may be drafted by an AI model from the measured check results only — the selection of issues, all scores, and all verification verdicts are computed mechanically, and the model cannot add issues or numbers. Ask the Data answers are AI-written from cached data and labeled. If the model is unavailable, built-in wording ships instead; grading never depends on it.

Public listings. Reviews of sites that are not publicly indexable — a noindex directive, robots.txt blocking all crawlers, or a sandbox/preview host (nip.io, vercel.app, ngrok and similar, or dev/staging subdomains) — are unlisted automatically: reachable by their direct link, but excluded from the directory, recent lists, search and the sitemap. This protects half-built sites from casual browsing.

Refresh cadence. Reviews run when a URL is submitted; there are no background re-crawls. Missing values are shown with the reason, never invented.