Skip to content
RadarWeb Español Join the waitlist

What we found on 12 websites

These are the figures from the audits we have carried out, without the name of any website. They are updated every time we approve a new report.

Last updated: October 18, 2026.

The main figures

Of the 12 websites we audited:

average score, out of 100
60
did not show up in any AI answer
33 %
take longer to load on mobile than Google recommends
92 %
do not measure visits to their website
33 %

Average score by area

Five scores from 0 to 100, using the same questions your customers would ask.

Each score starts at 100. An urgent issue takes off 20 points, an important one 8 and an improvement 2; no measured score goes below 20 points. If something could not be measured, the score says «not measured» and the report explains why.

SEO Google

Average score: 46 out of 100

SEO measures whether Google can read your website, where you rank when people search for what you offer and whether you show up on Google Maps. We check the code, links, security and structured data.

GEO ChatGPT and other AI

Average score: 77 out of 100

GEO measures whether ChatGPT, Gemini, Claude and Perplexity name your business when someone asks about what you sell without naming your brand, and whether they get your details right.

AEO Direct answers

Average score: 90 out of 100

AEO measures whether your website answers what people ask in a few lines, so Google can show it as a featured answer and AI assistants can quote it.

SXO Visitor experience

Average score: 34 out of 100

SXO measures what a person goes through on your website: mobile speed, accessibility, trust and privacy.

The AIO score has no figure yet: we publish it once at least 10 audits have measured it.

The most common issues

Each issue carries the title it has in the report. The figure says on how many websites it showed up, counting only the websites where that check could be done.

SEO · Google

Server security protections are missing

92 %11 of 12 websites

Why it matters
They stop, for example, another website from showing yours inside a frame to trick the visitor, or third-party code from loading. They are set up once and cost nothing.
How to fix it
Add them in the server or hosting settings. The code to copy is in the ZIP that comes with the report.

Photos without a description, marked as decoration

67 %8 of 12 websites

Why it matters
If they are photos with content (team, cases, services), Google Images and AI assistants do not know what they show.
How to fix it
Check each one: for those that show something, write an alt text that says what is in it.

There are titles that Google shows cut off

67 %8 of 12 websites

Why it matters
On Google, the result appears cut off with an ellipsis, and the part of the title that sets it apart is lost.
How to fix it
Keep each title between 50 and 60 characters, with what people search for at the start of the title.

There are descriptions that are too short or too long

58 %7 of 12 websites

Why it matters
A very short description does not explain what the page has, and a very long one appears cut off on Google.
How to fix it
Keep each description between 120 and 160 characters, saying what the person will find on that page. If the template repeats the same description, fix the template.

There are pages very far from the home page

58 %7 of 12 websites

Why it matters
Google reaches pages by following links from the home page. It visits the ones that are very far away less often and might consider them less important.
How to fix it
Bring those pages closer: link them from the home page, from the menu or from a category page one click away from the home page.

GEO · ChatGPT and other AI

The brand has no Wikidata entry

75 %9 of 12 websites

Why it matters
Wikidata is the public database Google and AI assistants check to know who is who. Without an entry, they rely on whatever they find scattered online and might confuse the brand with another one of the same name.
How to fix it
Put the brand details in order on the website and on its official profiles, so they all say the same. When there is press coverage or independent sources mentioning it, create the Wikidata entry.

There are very heavy pages

67 %8 of 12 websites

Why it matters
On the slow 4G network Google uses to measure (1.6 megabits per second), a 2,000 KB page takes about 10 seconds just to download, before it shows anything.
How to fix it
Remove repeated inline styles and scripts, and move the important text to the top of the HTML.

AI assistants are not naming the business

67 %8 of 12 websites

Why it matters
When someone asks an AI assistant who to turn to, the answer gives two or three names. If the business is not among them, that person does not reach its website.
How to fix it
Publish pages on the website that answer the questions customers ask, with concrete details (services, area served, prices if published), and complete the listings in the directories that AI assistants cite.

AEO · Direct answers

There are few direct answers in the text

80 %8 of 10 websites

Why it matters
A Princeton study (2023) measured that adding figures, quotes and sources to a text increases by up to 40% how often that text appears in AI answers. AI assistants quote paragraphs that answer a question on their own.
How to fix it
Start each section with a direct answer of 2 to 3 sentences, and then give the details.

There are pages with little text

70 %7 of 10 websites

Why it matters
With so little text, Google and AI assistants have nothing to tell them what the page offers, nor an answer they could quote.
How to fix it
Fill each page with what customers ask: what the service is, who it is for, how it works, timelines and prices, if they are published.

SXO · Visitor experience

There is low text contrast

92 %11 of 12 websites

Why it matters
People with low vision, or reading on a phone in the sun, might not be able to make out the text.
How to fix it
Darken the text or lighten the background until the contrast ratio is at least 4.5 to 1.

The website loads slowly on mobile

92 %11 of 12 websites

Why it matters
When loading goes from 1 to 3 seconds, the chance that the person leaves rises by 32% (Google and SOASTA, 2017). Google rates the page experience by how long the main content takes to appear (Google, web.dev).
How to fix it
If the main content is an image: compress it (WebP or AVIF), give it loading priority (fetchpriority="high") and defer the other images. If it is a block of text: check what delays it, such as blocking fonts (use font-display: swap), stylesheets and scripts in the head, and the server response time.

Cookies are set before asking for permission

83 %10 of 12 websites

Why it matters
In Chile, Law 21.719, which comes into force on December 1, 2026, will require a legal basis, usually consent, to process personal data. Tracking identifiers can count as personal data.
How to fix it
Set up the cookie notice so that analytics and advertising cookies are only activated after the visitor accepts.

There are forms without a privacy notice

75 %9 of 12 websites

Why it matters
People hand over their details without knowing what they will be used for.
How to fix it
Add a line next to the send button of each form with a link to the privacy policy. If the policy does not exist yet, publish it first.

There are other accessibility barriers

75 %9 of 12 websites

Why it matters
Each barrier shuts out a person who uses the website in a different way, for example with a screen reader or only with the keyboard.
How to fix it
Fix each item on the list following the instructions in the technical report.

Headings out of order

58 %7 of 12 websites

Why it matters
People who browse with a screen reader use headings as an index; when levels are skipped, that is lost.
How to fix it
Use the levels in order: H1, then H2, then H3.

What AI assistants answer

We asked each AI assistant the questions a customer would ask, without naming the website's brand. We count the website as showing up when the answer names the business or links to its website.

Answers in which the audited website shows up

  • ChatGPT Across the answers for 12 websites.12 %
  • Claude Across the answers for 12 websites.21 %
  • Gemini Across the answers for 11 websites.30 %
  • Perplexity Across the answers for 11 websites.18 %

Wrong details in AI answers

get wrong details from AI assistants, for example a phone number or a service that is not theirs
25 %

We compare what AI assistants say about each business with the details the owner gave us in the form.

Mobile speed

We measure the home page with Lighthouse in mobile mode. Google considers it a good load when the main content of the page (LCP) appears in 2.5 seconds or less.

for the main content of the home page to appear, the median across 12 websites
3.7 seconds
of home pages show it in 2.5 seconds or less, as Google recommends
17 %

This figure is different from the 92 % of websites that take longer to load on mobile than Google recommends. That other figure counts the home page, internal pages and real Chrome data; this one measures only the home page, with Lighthouse.

Trust and security

What a wary customer, a search engine or an AI assistant checks before recommending a website.

set cookies before asking the visitor for permission
83 %
have no privacy policy
50 %
lack the protection that forces the browser to always use a secure connection (HSTS)
50 %
have that protection with a period shorter than one year; the recommendation is at least one year
17 %

How we calculate it

The figures come from the audits we have approved. A person reviews each audit before it is sent. When a website has a second measurement, we only count the first one.

Each percentage is calculated over the websites where that check could be done. A figure is published only if it comes from at least 10 audits, and we never show the name of a website.

Speed is measured on the home page, with Lighthouse in mobile mode. For AI assistants we only count the questions that do not name the brand.

You can use and cite these figures naming RadarWeb as the source, under the CC BY 4.0 license.

Join the waitlistSee a sample report