Meikai AI Visibility Checker support

Health starts at 100 and loses 15 points per error, 5 per warning and 2 per notice, so a handful of errors moves the number a long way. The wording next to the ring tells you where you stand: Looking good from 85, Needs attention from 60, and Critical issues below that.

The chips under the ring score the page by area: Crawl, SEO, Read (readability), A11y (accessibility), and GEO. GEO appears on product pages only. Readability uses Flesch reading ease, which is calibrated for English, so it is left out rather than scored on other languages.

The first screen lists the three most severe issues. Everything else is in the All issues panel, ordered errors first.

An audit is kept in Chrome session storage for the tab and reused for about ten minutes, so reopening the popup is instant. Re-run overrides the cache. The cache is dropped when the browser session ends and when the extension updates.

Reads robots.txt from the site and resolves allow or deny for the current path and for the site root. Training crawlers and retrieval fetchers are listed separately, because sites routinely block one and allow the other:

• OpenAI: GPTBot (training), OAI-SearchBot, ChatGPT-User
• Anthropic: ClaudeBot (training), Claude-SearchBot
• Google-Extended
• PerplexityBot

The line below the list reports whether robots.txt was found, plus any meta robots or X-Robots-Tag directive that affects indexing. A file that was never reachable is reported as not readable, which is deliberately not the same as missing.

The same panel holds the optional GPTBot live check. Once you grant access for the site, the extension requests the page once with the GPTBot user agent and reports the HTTP status and whether the response was blocked. That is how you catch a CDN or WAF rule that robots.txt says nothing about. Until then the check reads as optional and nothing else is affected.

Word count, HTML weight with a flag past 2 MB, readability on English pages, and the most frequent one, two and three-word phrases with navigation and boilerplate filtered out. A thin page or a phrase list that has nothing to do with the topic is usually the first sign that an assistant will summarise the page badly.

Probes the site root for the machine-readable files agents look for, and reports whether /llms.txt is a real map or a stub, alongside /llms-full.txt and /AGENTS.md. When /llms.txt exists, the panel also shows its title, whether it has a summary, and how many sections and links it contains. /llm.txt is probed as an alternative filename some sites use. These files matter for agent browsing; they are not a Google Search ranking factor.

Compares the server response with the rendered page to show what a crawler gets without running JavaScript: HTTP status and response time, raw against rendered word count, and whether the title and JSON-LD are present in the HTML. The verdict separates a page that merely hydrates from a shell that has no content at all, names the hydration payload when it can identify one, recognises a Markdown response served to crawlers, and calls out a bot wall as a different document rather than a rendering problem.

• Title, meta description, and canonical, including whether the canonical points at this page
• H1 count and heading structure
• Indexability from meta robots and X-Robots-Tag together
• HTTPS and images with no alt text
• The schema.org types found on the page, and JSON-LD that fails to parse

This panel opens only on pages the extension classifies as product pages, from the URL and from Product schema. It checks what a shopping agent needs to quote the product with confidence:

• Product and Offer fields: name, price, currency, availability, brand, SKU or GTIN, image, condition
• Offer completeness across variants
• Whether the product URL in schema agrees with the canonical
• Answer-first copy, FAQ or Q&A blocks, and breadcrumb schema
• Fact density, and whether the visible price and rating match the structured data

Every finding with its severity, errors first. Codes and wording match the Site Scanner catalogue, so a fix that clears an issue here clears it in a full scan.

Open Meikai goes to the Meikai app sign-in and Full Site Scanner to the Site Scanner page, both carrying the audited URL so you can continue there. Neither is needed to run or read an audit.

Reload the tab and open the popup again; Try again in the error state repeats the audit. A page that Chrome protects from extensions, such as a chrome:// page or the Web Store, cannot be audited at all. If the message mentions an outdated build, reload the extension from chrome://extensions or reinstall it from the Web Store.

Check which wording you got. Missing means the site answered and the file is not there. Not readable means the request never completed, which a proxy, a strict content security policy, or a dropped connection can cause. Re-run before you change anything in production robots rules.

Click Enable in the AI access panel and accept the Chrome prompt; the audit re-runs by itself once permission is granted. If no prompt appears, Chrome may be blocking permission requests for that site, in which case grant access from chrome://extensions instead.

The page was not classified as a product page. If it is one, send us the URL and whether Product schema is present in the source, and we will look at the classification.

From the Chrome Web Store: search for Meikai AI Visibility Checker, or open the listing linked on this page.

Yes. The audit is free and does not require a Meikai account or a trial.

No. The audit is computed in the browser from the tab you opened and the public files on that site. The Open Meikai and Full Site Scanner buttons are the only things that navigate to Meikai, and only when you click them.

Google Chrome on desktop. Other Chromium browsers generally load the extension, but we only test and support Chrome.

Yes, as long as the page is served over http or https and your browser can reach it. Checks that read robots.txt or agent files need that origin to answer the request, so a locked-down environment may report them as not readable.

The extension grades one page from what your browser can see. Site Scanner crawls the whole site, renders pages remotely, keeps history, and aggregates scores across properties, so its view is broader.