How we score
How we score your website
Kashif reads your website and gives it a score from 0 to 100, with a list of what to fix first. This page says what we read, what we leave out and how the number is built.
The short version: we read your site. We do not ask any assistant about you.
What we read
We read the public pages of your site, the way a search engine’s crawler does. Our crawler is called KashifBot.
Your site’s files for crawlers.
robots.txt first, then the sitemap, llms.txt and a file for AI agents, if they exist.
Your pages.
Up to 8 pages in a full check, or up to 3 in a quick check, starting from your homepage.
Your speed.
Google PageSpeed Insights measures it.
Screenshots.
In a full check, of your homepage and one contact page.
We read what any visitor can see. We do not log in, fill in forms or make payments. We change nothing on your site.
We read each page as your server sends it. Text that only appears after a script runs may be missed.
If your robots.txt asks KashifBot to stay out, it stays out, and those checks show as not measured. The speed test and the screenshots do not run as KashifBot, so those rules do not cover them.
Except on the free quick scan, we also look up your domain's public records. Those lookups go to DNS servers and the domain registry, not to your site, and they do not count in your score.
Read more about KashifBotThe six areas and the comparison
Six areas make up your score. These are the names you will see in your report.
Getting found
Can search engines read your pages, and do they say what you sell?
We check: that your site opens and its security certificate is complete. Page names and descriptions. One clear headline on each page. A sitemap. Nothing telling search engines to stay out. Language tags. Business details in a form search engines can read. How a link to your homepage looks when shared.
Not here: your position in Google.
Readable by AI
Can an assistant read the facts on your site?
We check: that your robots.txt lets AI crawlers in. Facts in a form assistants can quote. Enough text to answer a customer's question. Your business name on your homepage. For restaurants and cafes, whether the menu is text or a picture. A short guide file for AI (llms.txt).
Not here: what any assistant does with those facts.
The first impression
Does your site load fast and say what you do?
We check: speed on a phone, how soon the page appears, and whether the layout jumps. Speed on a computer. In a full check, the AI model also looks at screenshots of your homepage.
Not here: how your real visitors feel about the site.
Trust at a glance
Does your site look and read like a business a customer would trust?
We check: whether your homepage shows a picture when someone shares the link. In a full check, the AI model reads your text and looks at your homepage for a clear offer, language and a consistent look.
Most of this area is the model's judgement, so read it as an opinion.
Not here: whether your business is licensed or honest.
Easy to reach
How many taps from your site to a person?
We check: a WhatsApp chat link. A number that can be tapped to call. A contact form. An email link. How easily a customer finds your contact details. Links to your social accounts.
Not here: whether anyone answers.
Ready for AI agents
Can an AI agent read and use your pages?
We check: page structure, such as image descriptions and named form fields. Whether the layout holds still. Whether you publish a file of actions an agent can call. No such file does not count against you.
This area uses rule-based checks only. No AI model marks it.
Not here: we do not test whether an agent can finish a booking or an order.
Next to your competitors
On Pro and Premium, and during the free trial, we also compare you with competitor sites you add. We run many of the same rule-based checks on each site (not speed), on up to 3 pages, and show where they are ahead.
This comparison sits beside your score in your account. It does not change the number.
How the score is built
Each check gets a result.
Pass, Warning, Missing or N/A. A pass earns full points, a warning earns part, and Missing earns none. N/A is left out, so it can neither help nor hurt you. Some checks count for more than others.
Each area gets a mark from 0 to 100.
In a quick check, the mark comes from the rule-based checks alone. In a full check, an AI model also reads your pages and gives its own mark for five areas: Getting found, Readable by AI, The first impression, Trust at a glance and Easy to reach. The two marks are blended, and the model's counts for more. Ready for AI agents stays rule-based.
The areas become one number.
They do not count equally. Easy to reach counts most. Ready for AI agents counts least for now, because it is a new standard.
What we could not measure stays out.
An unmeasured check or area is not scored as zero. It is left out, the rest is scaled to fill the gap, and your report names what is missing.
We explain the method here, not the exact weights.
What the score means
Your score is a snapshot of what we could see on the day. A higher score means more of what we check is in place.
It is not a ranking, a forecast of sales or a grade of your business.
We don't promise rankings or recommendations. Nobody honestly can. Google says the same:
“No one can guarantee a #1 ranking on Google.”
We show what we found and what we saw, and check whether the fix worked.
The bands
| Score | Band | What it means |
|---|---|---|
| 0 to 39 | Weak | Serious gaps. Start with the first fix. |
| 40 to 69 | Fair | Some of it is in place, and some is missing. |
| 70 to 100 | Strong | Most of what we check is in place. |
These lines are our own. They are not an average of other sites, and we do not publish one.
If too much could not be measured, you see "provisional — limited data" in place of a band word. A quick check is always marked this way.
Quick check and full check
| Quick check | Full check | |
|---|---|---|
| Time | About a minute | A few minutes |
| Pages | Up to 3 | Up to 8 |
| Rule-based checks | Yes | Yes |
| Speed test | Phone only | Phone and computer |
| Screenshots | No | Yes |
| AI model reads your pages | No | Yes |
| You get | A provisional score | A score, findings and a fix list |
The free scan starts with a quick check, so you see a number in about a minute. A full check then follows, unless we have reached our daily limit. When it finishes, its score replaces the quick one on the page.
A full score can land well below the quick one. The quick check only asks whether things are there. The full check also reads them.
Do not compare a quick score with a full score. We keep them apart for that reason.
What your score leaves out
Your score does not include:
- Your place in Google or any other search results
- What any assistant tells people
- Your visitors, calls, orders or sales
- Your listings, reviews and ratings on other sites
- Your social media posts or followers
- Anything behind a login, or pages past the limit above
- Whether what your site says is true
It is not a security test or a legal review.
Some things we notice do not count. In your account they sit under "Worth a look (not scored)", such as broken links or page names that run long. They do not change your score.
We do not ask assistants
We don't ask ChatGPT or any assistant about you. We never put a customer's question to an assistant to see what comes back.
We check what assistants and search engines can read on your site. What an assistant says changes from one question to the next, so we measure what you can control.
So a strong Readable by AI score means your facts can be read. It does not mean any assistant will name your business.
How we use AI
An AI model helps write your findings. In a full check it also gives the marks described above, so it affects your score.
It works only from what we collected: the text of your pages, the results of our checks and, for The first impression and Trust at a glance, screenshots of your homepage.
We tell it to base every finding on something in that data, and not to invent traffic, sales or rankings. We do not ask it what it knows about your business.
A model can still get things wrong. Each finding shows what we saw, so you can check it.
The rule-based checks use no AI. If the model is not available, the area marks come from those checks alone, and your report says "Not reviewed this time".
Re-checks
A score carries a date. Your site changes, so we check again.
It depends on the plan. Free: no automatic re-checks. Pro: a full audit re-check every two weeks. Premium: a full audit re-check every week.
You can also run a check from your account. When a fix is done, "Check again" re-reads that one check. Speed is the exception: it waits for your next check.
Your score can move even when your site has not. Speed varies from one test to the next, the model's read is a judgement, and we sometimes improve how scoring works.
What to do with your result
Start at the top.
Under "What to fix first", anything critical comes first. After that, the fixes that matter most for the least work come first.
Read the finding.
Each one says what we saw, what it costs and what to do. "What it costs" explains how the problem loses customers. It is not a number we measured.
Fix it, or pass it on.
Do the easy ones yourself, or hand the list to whoever runs your site. On WordPress, Wix, Shopify, Squarespace, Salla and Zid, some findings show where to click.
Check it worked.
Use "Check again", or wait for your next check.
Stuck?
Message us. A person answers.