See how your site looks to Google and to AI.
Enter an address and the tool runs more than fifty concrete checks — from Core Web Vitals through structured data to AI bot rules. No generic advice from a language model, just measurable facts and pre-written recommendations.
The audit works with the downloaded HTML. On sites that render content with JavaScript (SPAs), a static fetch may not see everything. The speed section uses Google PageSpeed Insights — without your own API key a lower request limit applies.
Speed
25%Core Web Vitals from PageSpeed Insights, measured separately for mobile and desktop.
SEO
25%Indexability basics and the on-page signals that decide your position in search.
AEO
12%How ready the content is for a direct answer to be extracted from it.
GEO
13%Visibility for generative AI and rules for the bots that may cite you.
Soundness
13%Technical health and reliability — the things that quietly spoil both browsing and crawling.
Accessibility
12%Usability for keyboards and screen readers — increasingly a legal requirement too.
How it works
Four steps, no black box
The whole audit is a sequence of concrete measurements and rules. Every number in the report traces back to the check it came from.
- 01
We fetch the HTML
We load the source of the given address and measure the server response along the way. No browser is launched, so it takes a fraction of a second.
- 02
We read the companion files
In parallel we read robots.txt, llms.txt and the sitemap, and verify up to fifteen internal links. If a source drops out, the audit finishes without it.
- 03
We measure speed at Google
PageSpeed Insights measures Core Web Vitals for both mobile and desktop. The two runs go in parallel, each in its own request, and are retried on failure.
- 04
We score it by rules
Every check has a status and a weight. The recommendations are pre-written and picked by the measured value — nothing is generated by a language model.
What we check
Over fifty checks across six areas
The weight on each category is the share it contributes to the overall score. Speed is measured separately for mobile and desktop, together making up 25%.
Speed
25% of scoreCore Web Vitals from PageSpeed Insights, measured separately for mobile and desktop.
- LCP
- INP / TBT
- CLS
- TTFB
- Performance score
SEO
25% of scoreIndexability basics and the on-page signals that decide your position in search.
- noindex
- Title
- H1
- Canonical
- Open Graph
- robots.txt
- Sitemap
AEO
12% of scoreHow ready the content is for a direct answer to be extracted from it.
- FAQ / HowTo schema
- Question headings
- Direct answer
- Paragraph length
GEO
13% of scoreVisibility for generative AI and rules for the bots that may cite you.
- llms.txt
- AI bots
- Entity profiles
- Contact & author
- Content freshness
Soundness
13% of scoreTechnical health and reliability — the things that quietly spoil both browsing and crawling.
- Viewport
- Favicon
- Custom 404
- HSTS
- Broken links
Accessibility
12% of scoreUsability for keyboards and screen readers — increasingly a legal requirement too.
- Landmarks
- Form labels
- Accessible names
- Tab order
General advice
What to do with the result first
Advice that holds across sites, whatever your audit says. Recommendations for individual findings are inside each check in the report.
Fatal findings first, numbers second
If the report shows a red band at the top, tuning LCP is pointless. A noindex directive or a robots.txt ban means the site simply is not in search — and while that holds, nobody sees any other optimisation.
One image often decides it
On most sites the main brake is a single large image above the fold. Shrinking it, converting to WebP or AVIF and giving it fixed dimensions usually fixes LCP and CLS at once.
Answer first, context second
Put one or two sentences that directly answer the question under each heading, and only then elaborate. This pattern is the basis of getting cited by assistants.
Separate training from citation
Blocking GPTBot or ClaudeBot while letting OAI-SearchBot, Claude-SearchBot and PerplexityBot through is a sensible combination. A blanket Disallow cuts you off from citations too.
Run the audit twice
Lighthouse measures slightly differently every time and a ten-point gap between runs is normal. Before you fix anything, confirm the finding is not just measurement noise.
The audit sees one page
Exactly the address you enter is evaluated — not the whole site. Run the homepage, a typical product page and one article separately; their findings usually differ.
Guides
What people come here to find out
The five questions we get most often. Each has a page that answers it in detail — what gets measured, how to read the result and what to do next.
Frequently asked
What people ask most
If something is missing, the answer is often in the report itself — every check has its own explanation.
AEO (Answer Engine Optimization) is how ready a page is for a direct answer to be extracted from it — FAQ schema, question-style headings, short paragraphs and structured content. GEO (Generative Engine Optimization) covers visibility for generative AI: whether the site has an llms.txt file, whether robots.txt lets search bots such as OAI-SearchBot or PerplexityBot through, and whether it is possible to verify who is behind the site.
Every check has a status (fine / could be better / missing) and a weight based on severity. A category score is their weighted average, and the overall score is a weighted average of the categories — SEO 25%, speed 25% combined, GEO and Soundness 13% each, AEO and Accessibility 12% each. Measurable metrics such as LCP or CLS are not converted in hard steps but continuously, so even a small improvement shows up.
When too few checks in a category can be verified (typically when a PageSpeed Insights measurement does not finish), a dash appears. A number computed from the two remaining checks would look trustworthy while saying nothing. Such a category is left out of the overall score, and the report offers a button to run the missing measurement.
The results routinely differ by tens of points — a page can be brisk on desktop and crawl on a phone. Google also indexes mobile-first, so the mobile value carries more weight. The browser requests both measurements separately and runs them in parallel, because Lighthouse at Google is the slowest part of the whole audit. The report appears only once both results are in.
A noindex directive, a site-wide crawl ban in robots.txt, missing HTTPS or an error response from the server all mean the site is effectively absent from search. Without a cap such a site could still score 90 because the other checks pass. The score is therefore capped at 35 points and a red warning appears above the report.
No. The audit works with the downloaded HTML and does not launch a browser. On sites that generate content in the browser (SPAs) it may not see the text, headings or links. If a page looks that way, the report flags it in the measurement notes. The speed section is unaffected — Google measures that with its own browser.
No. All the text is pre-written and selected by the outcome of a specific check. When you see a recommendation for LCP, it is the same text every site in the same value range receives. That makes the advice consistent, verifiable, and impossible for the tool to invent.
No. Nothing is saved to a database or a history — the audit is recomputed on every request and the result exists only in your browser. If you want to keep a report, save the page or print it to PDF.
All that is left is the address.
The audit takes about a minute, a little longer on slower sites, and nothing is stored. Feel free to run it on a competitor — all you need is a public URL.
Run an audit