Technical Audit
Audit your site's foundation (indexing settings, page HTML, broken links, load speed, backlinks, competitors)
Not the articles, the foundation
"Sorabun > Measure / Analyze > Technical Audit" audits your site's foundation rather than the content of your articles (Pro).
Everything checked here shares one trait: the screen still looks fine even when something is broken. The classic example is forgetting to turn off "discourage search engines from indexing this site." Leave that setting on and no amount of writing gets you into search results. The site displays normally, so months can pass before anyone notices.

What does it check?
Click "Run Audit" and Sorabun inspects the following within a few seconds, ordered from the heaviest issue (needs attention) down.
| Item | What it looks at |
|---|---|
| Search engine visibility | Whether the setting to block indexing has been left on |
| Permalink structure | Whether the URL structure is still stuck on ?p=123 |
| HTTPS | Whether the site URL is still on http |
| Sitemap | Whether WordPress itself, an SEO plugin or the Sorabun XML Sitemap is serving one |
| robots.txt | Whether it contains anything that blocks all crawlers |
| AI search crawlers | Whether robots.txt blocks OAI-SearchBot, Claude-SearchBot, Claude-User, or PerplexityBot. Blocking them makes the site less likely to appear in that AI's search answers (per each vendor). Training crawlers (GPTBot, ClaudeBot, Google-Extended and so on) are separate, so blocking them does not count against you |
| Snippets | Whether the robots meta tag or X-Robots-Tag on the home page and the latest article contains nosnippet or max-snippet:0. Google says a page must be eligible to show a snippet to appear in AI Overviews and AI Mode |
| Overlapping structured data | Whether the HTML of the latest article contains two or more Article structured data nodes. Sorabun merges into one with Yoast SEO, Rank Math, and All in One SEO, but cannot merge what a theme or another plugin outputs, so it counts the real output and tells you |
| Fetching as an AI search crawler | Whether the latest article can actually be fetched with the user agents of OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User, and PerplexityBot. Even when robots.txt allows them, a hosting, CDN or WAF setting that blocks AI crawlers, or a security plugin, may be returning 403. The request is made from this server itself, so a block based on the connecting IP address cannot be detected |
| Broken internal links | Internal links with no destination (typically links to a deleted article) |
| Duplicate titles | Whether articles with the same title are competing against each other for ranking |
| Mixed content | Whether an https site still has images or other assets on http:// |
| Image alt attributes | Images with no alt text |
| Title length | Whether the title is long enough to get cut off in search results |
| Orphaned articles | Articles with no internal links pointing to them |
The body scan covers up to 200 of your newest articles. The number scanned is shown with the results.
Any item that finds a broken link or an image without alt text lists a link to the article in question. Open it, fix it, then run the audit again.
Page HTML check
The table above reads the article bodies stored in the database. Search engines, however, read the finished page that your theme and SEO plugin put together. The title, meta description, canonical and noindex all live in the <head>, and no amount of looking at the article body will reveal them. Click "Read pages" and Sorabun reads your own pages the way an outside visitor would, without logging in (no login cookies are sent).
- Which URLs: the home page, published posts and pages (newest first), a few category and tag archives, and the URLs in your sitemap. Up to 500 in total. It does not crawl by following links
- How it reads: to keep the load on your site light, WP-Cron reads about 30 pages at a time and waits 2 minutes before the next batch. If several pages in a row come back as not found (4xx), it stops that batch early so your own server never trips the "repeated 404s" lockout of the Sorabun firewall. It keeps going after you close the screen
- Cost: none (it only reads your own site)
| Item | What it looks at | Level |
|---|---|---|
| Server error (5xx) | The page returns a 500-range error | Needs attention |
| No title | There is no <title> (a title inside an SVG does not count) | Needs attention |
| Redirect loop | Following the redirects leads back to the starting URL | Needs attention |
| Kept out of search (noindex) | The post itself is set to appear in search, but the robots meta tag or X-Robots-Tag says noindex (needs attention on the home page) | Caution |
| Page not found (4xx) | A URL Sorabun knows about, including sitemap URLs, returns 404 or similar | Caution |
| In the sitemap but noindex | A page you deliberately keep out of search is listed in the sitemap | Caution |
| Redirect chain | It takes two or more redirects to reach the page. A chain is counted once, at its first URL | Caution |
| Conflicting canonical | The canonical in the HTML and the canonical in the HTTP Link header point to different URLs | Caution |
| Duplicate title, description or body | Two or more pages share the same one. Bodies are compared by a hash of the text in the article body area | Caution |
| No meta description | There is no description for search results | Caution |
| No H1, more than one H1 | Also tells whether the extra H1 is in the body or outside it (a theme that marks up the logo as H1, for example) | Caution |
| Thin content | Fewer than 400 characters in Japanese, or 150 words in English | Caution |
| Title too short, description too short or too long | See "Length guidelines" below | Info |
| Skipped heading level | An H4 right after an H2, for example. Tells apart the article body and theme areas such as the sidebar | Info |
| Canonical points to another URL | The page names a different URL as the official one (fine if intended) | Info |
| Slow response | The HTML took more than 1.5 seconds to come back | Info |
| Could not read | 429, 401, 403, a bot check page, or no connection | Info |
The items in the table above (AI crawlers, orphaned articles, broken internal links, mixed content, image alt text, and the length of the titles stored in the database) are not checked again here.
Length guidelines (adjusted for Japanese)
| What | Japanese | English and others | Why |
|---|---|---|---|
| Title too short | Under 8 characters | Under 10 characters | A title like "Company" or "Blog" only says what kind of page it is. One Japanese character carries about as much width and meaning as two English letters |
| Description too short | Under 40 characters | Under 70 characters | Mobile search results show roughly 50 to 80 Japanese characters. A short description leaves the space unused, and Google is more likely to swap in text from the page |
| Description too long | Over 120 characters | Over 160 characters | Even on desktop, it gets cut off at around 120 Japanese characters |
| Thin content | Under 400 characters | Under 150 words | Japanese has no spaces between words, so it is counted in characters. Only the article body area is counted, not the menu, sidebar or footer |
The Japanese guidelines apply when at least 30% of the text is kanji or kana.
What it does not call broken
- 429 (rate limited) and blocked pages (401, 403, bot check pages) are listed as "Could not read", not as missing pages. A security plugin or WAF may be blocking reads from your own server
- Pages you deliberately keep out of search (posts set to noindex in your SEO plugin, archives, attachment pages, search results) are not flagged for noindex. They are flagged only when they are also listed in the sitemap
- noindex pages and pages whose canonical points elsewhere are left out of the duplicate checks, because their owner has already dealt with the duplication
- When the whole site is set to "Discourage search engines from indexing this site", per-page noindex is not reported (the "Search engine visibility" row in the table above covers it)
Where to fix it
Each flagged page comes with where to fix it. The same kind of issue can need a fix in different places depending on the page.
- Theme: no title, no H1, a logo marked up as H1, skipped heading levels in the sidebar
- SEO plugin: the title, description, noindex and canonical of the home page and archives (site-wide settings)
- Post settings: a post's title, description and canonical (the SEO plugin box on the edit screen). Use the "Edit" link to open it
- Article body: an H1 inside the body, skipped heading levels in the body, thin content, duplicate bodies
- Server, redirect settings, sitemap, security settings: 5xx errors and slow responses, redirect chains and loops, missing URLs in the sitemap, blocked pages
Changes since last time
After the first read, Sorabun reads the pages again every week and shows new issues and fixed issues since the previous run. An issue counts as fixed only when the page was read again and the issue is gone (pages that are no longer read because you deleted the post do not count). The weekly re-read can be turned off under "Settings > Generation defaults > Page HTML check re-read" (pagecheck_auto, on by default). On a site where no one has clicked "Read pages" yet, Sorabun never starts reading on its own.
IndexNow key
When you use IndexNow, search engines fetch your site's verification key before accepting the URLs you send. Sorabun does not store the key as a file; WordPress serves it. If a server redirect (with or without www, http to https) or a cache answers with something else first, the key cannot be read. Click "Check the key" and Sorabun checks both the domain and the www variant for a missing key, wrong contents, or a redirect (it also checks after each page read finishes). A redirect from the www variant to your site's own key is the normal setup, so it is not flagged.
The choice of checks and how they are judged (not calling blocked pages broken, counting a chain once at its start, leaving noindex and canonicalized pages out of duplicate checks) are based on the open source SEO tool OpenSEO (MIT License, (c) 2026 Ben Senescu), rebuilt to run inside WordPress. The length guidelines were set again for Japanese.
Lighthouse scores (several pages)
Sorabun runs the home page and the newest item of each post type (posts, pages and custom post types, skipping any set to noindex per post) through PageSpeed Insights on both mobile and desktop, and lists the Performance, Accessibility, Best Practices and SEO scores. 90 and above is green, under 50 is red.
- There is no cost (PageSpeed Insights is free). It only measures when you click the button, and one run makes "number of pages × 2" calls
- Up to 30 calls a day. Without a PageSpeed Insights API key (
psi_api_key), you hit Google's own limit sooner - Google's servers fetch the pages, so a site that cannot be reached from outside (such as a local development site) cannot be measured
- Core Web Vitals below are real measurements of the home page (your visitors' numbers); these are lab scores
Core Web Vitals (load speed)
Sorabun measures the load speed metrics Google uses to judge rankings, on your top page (mobile), using PageSpeed Insights.
- LCP: time until the largest element on the page appears. Up to 2.5 seconds is "good"
- INP: responsiveness to clicks and taps. Up to 200ms is "good"
- CLS: how much the layout shifts. Up to 0.1 is "good"
If real visitor data (CrUX) is available, Sorabun uses it. On sites with too few visits, it falls back to lab data instead, and the screen shows which one was used.
It works without an API key, within a rate limit. If measurements start failing, set up "Settings > Search Data Source > PageSpeed Insights API Key" (psi_api_key), which Google Cloud issues for free.
Backlinks (how the outside world rates you)
SEO rests on three pillars: content, technical health, and authority. Backlinks are that third pillar. Using the same DataForSEO credentials you set up for search volume, Sorabun shows you:
- Referring domains: how many different sites link to you. Check this before the raw link count
- Backlink count: the total, including multiple links from the same site
- Anchor text: what words are used to link to you. Mostly your site name means you're known by brand, and a mix of the keywords you're targeting means those words are gaining authority too
- One-year history: referring domains and backlinks month by month, plus the sites gained and lost that month. A slow decline means the pages that linked to you are disappearing
- New and lost links: everything from your previous fetch up to today (30 to 90 days). Only the strongest link from each site is shown. If a strong site's link is listed as lost, the page may simply have moved, so open the linking page to check
- Referring sites: the 100 strongest. Sites with a DataForSEO spam score of 40 or higher are hidden by default. Tick "Include likely spam sites" before fetching to show them too
Fetching all of this costs about $0.13 per click on the paid API, so it only refreshes when you click the button. Once a month or so is plenty. Anything fetched in the last 12 hours is not fetched again (3 days for the history), so clicking twice costs nothing extra. Under the button you'll see what a click would cost right now and how much you've spent today. Sorabun remembers the previous numbers, so you can also see whether you gained or lost ground.
Buying backlinks or trading them excessively violates Google's guidelines and can trigger a penalty. This screen exists to show you where you stand, not to offer a shortcut for growing the count. Writing original material that people want to cite (research, data, real examples) looks like the slow path, but it's the reliable one.
Competitors
Competitor domain strength
A domain strength number means nothing on its own if all you look at is your own site. You need something to compare it against.
This screen lets you compare yourself against the domains that actually show up on page 1 for the keywords you're tracking. You don't need to type in any competitors. Sorabun records which domains were on top from the search results Rank Tracking already fetches every day, so this adds no extra search cost since it reuses that same response.
Click "Fetch competitor strength" and Sorabun fetches the recorded domains plus your own site in a single call. Your own site goes into that same call because fetching it separately would put the two readings at different points in time, and you couldn't compare them.
There's just one thing to remember when reading the table.
- Keywords where you rank below a weaker domain: these are winnable on article quality alone, and fixing them is the fastest path forward.
- Keywords where every domain on page 1 is far stronger than you: going after these head on won't get you anywhere right now. It can be worth deferring them in favor of more specific keywords instead.
This is also a paid API (a few yen per call), so it only fetches when you click the button.
0 to 1000 is DataForSEO's own metric, not a Google metric. Google has repeatedly said it doesn't use anything like a domain authority score. Don't read the absolute value. Look at the gap to your rivals and your own trend over time instead.
Competitor keywords and top pages
Knowing how strong a rival is doesn't tell you which keywords it ranks for. Click "Look up" in the table, or type a domain and click "Look up keywords", and Sorabun checks that domain on Google Japan in Japanese:
- Estimated monthly traffic and the number of ranking keywords: roughly how many people it gets from search
- Ranking keywords: up to 100 keywords in the top 20, by traffic. Each shows the position (among organic results, not counting ads or "People also ask"), monthly searches, difficulty (0 to 100), search intent and the page
- Top pages by traffic: 10 pages
Each domain costs about $0.05. A domain looked up in the last 3 days is not fetched again. Sorabun keeps up to 12 domains and drops the oldest when you go over. Results you no longer need can be removed with "Remove this result".
A big portal (shopping, encyclopedias, video) gets most of its traffic from keywords about its own business, which have nothing to do with you. Pick sites in the same field that compete for the same readers.
Keywords only your competitors rank for
Sorabun lists the keywords where a competitor you looked up is in the top 10 and your site isn't, with the most searched first. Your own positions come from the Rank Tracking table (measured positions and the Search Console positions recorded daily) and the last 28 days of Search Console queries read through Site Kit. Differences in case, full-width and half-width characters, and spaces are ignored, so "WordPress 高速化" and "wordpress高速化" count as the same keyword. Keywords where your site ranks 11th or lower stay in the list with your position shown.
Keywords containing a competitor's name (brand searches such as "Acme pricing") are left out, since no article of yours can win them.
Click "To article ideas" to send a keyword to "What to write next" in SEO Insights. Sent keywords survive the weekly automatic refresh, and you can create an article from there.
Cost and the daily limit
Backlinks, competitor strength and competitor keywords are all fetched only when you click. Nothing is fetched automatically, and only administrators can click.
To avoid overspending, "Settings > Search Data Source > Competitor and backlink fetches (daily limit)" (research_daily_usd, $1 by default) sets how much may be spent per day. When a click would go over, Sorabun tells you instead of fetching. Set it to 0 to stop fetching altogether.
The referring sites list, history, new and lost links, and the competitor keyword lookup are based on the design of the open-source SEO tool OpenSEO (MIT License, (c) 2026 Ben Senescu), rebuilt to run inside WordPress.