NowThe AI Readiness StudyPart 1 of 4
Is your website ready for AI search? Only 6 of 1,000 brand websites passed every check
Only 6 of 1,000 brand websites passed every check. The 12 basics that help AI read a site, how many brands have each, and the FAQ most of them miss.
Gabriele FajaFounder, remake
· 7 min read
It’s Monday, and a screenshot lands in your inbox (from a client, or from your boss). Someone asked an AI assistant about your category, and a rival came back instead. Before anyone rewrites the brand story, ask a duller question. Can an assistant read your website, and does it answer what buyers ask?
AI assistants read plenty of pages about your brand. Your own website is the only one you fully control. So we read 1,000 brand websites and ran each through our website check. It looks for 12 basics that help AI reach and read a site, plus some smaller extras. Only six passed every check without a single warning, and they come from all over the map: a campsite and a print shop, a software company and an agency, an online doctor service and a group of care homes. (Whatever you sell, your site can join them.)
| Passed every check | 6 of 1,000 |
|---|---|
| All 12 basics, with warnings | 151 of 1,000 |
| Missing only an FAQ | 122 of 1,000 |
| The rest: another basic missing | 721 of 1,000 |
Each square is one site. The 6 and the 151 in pale teal have all 12 basics; the 122 in teal miss only an FAQ.Source: remake study of 1,000 brand websites read 14–15 Oct 2026. Company list from Wikidata (CC0).
How to be seen for AI search?
Exactly what a visitor needs (AI assistants included). The oldest trick in the world! Who you are, what you do, and how to reach you, all in plain text.
Run your eye down the bars. The technical basics hold up well (at least 924 sites on each of the first seven bars). Then the numbers thin out. Fewer than half have code saying who runs the site, and fewer still have an FAQ.
| Text that shows without running scripts | 990 of 1,000 |
|---|---|
| AI crawlers free to read the site | 979 of 1,000 |
| A homepage that loads without a long wait | 972 of 1,000 |
| A page title | 966 of 1,000 |
| A secure connection | 962 of 1,000 |
| A layout set up for phones | 953 of 1,000 |
| At least a few hundred words of text | 924 of 1,000 |
| A phone number or email address | 824 of 1,000 |
| A page description for search results | 797 of 1,000 |
| A main heading on the homepage | 695 of 1,000 |
| Code that says who runs the site | 457 of 1,000 |
| An FAQ page or section (96 also mark it up in code) | 415 of 1,000 |
Each bar is out of the 1,000 brand websites we read in full.Source: remake study of 1,000 brand websites read 14–15 Oct 2026. Company list from Wikidata (CC0).
What do the best brand websites do?
We got to know three of them well. They have nothing in common, and that’s the point. Each one does something any brand can copy.
- Camping Armanello, a family campsite in Benidorm, wants you there in winter. It doesn’t ask you to take its word for it. “No lo decimos nosotros: lo dicen los datos,” says the homepage (“we’re not saying it, the data is”). Then come five winters of readings from AEMET, Spain’s state weather agency. No significant rain on 89% of days from November to March, and an average high of 18.9°C. (Hard to argue with a weather station.)
- WIRmachenDRUCK, an online print shop in Backnang, Germany, leaves a business card in its homepage code. Visitors never see it. Software asking who runs the site finds the company’s full name, its street address and a link to its Wikipedia article.
- Benevity, a Canadian company, makes workplace giving software. Its homepage opens with a slogan (“Trust at scale. Undeniably proven.”) and right after, it says what it sells. “The Benevity® Enterprise Impact Platform is one trusted platform for employee giving, volunteering, grants and reporting.” The slogan is for the heart. The sentence is for anyone who has to explain the company to someone else.
Do you need an FAQ page for AI search?
Before anyone buys, they have small, practical questions. What does it cost? When will it arrive? Are there nuts in it? Can I bring the dog? Now they can just ask an AI assistant. An FAQ is where you write the answers down, in your own words.
Where the answers sit matters too. Picture your delivery times folded away behind a little plus sign. A person clicks. Software might not. Microsoft’s guide to appearing in AI answers says it plainly: “AI systems may not render hidden content.” So keep every answer open on the page.
How much an FAQ sets you apart depends on your neighbours. For a bank or an insurer, it’s keeping up. In food and drink, it puts you ahead of most of the category. A few to start with:
- Food and drink: Is it gluten-free? Where can I buy it? How long does it keep once it’s open?
- Finance and insurance: Can I apply? How long does approval take? How do I make a claim?
- Online shops: Do you deliver to me? When will it arrive? How do I send it back?
- Travel and hospitality: What time is check-in? Is there parking? How far is it to the centre?
| Finance and insurance brands (60 of 91) | 66% ±10 |
|---|---|
| All 1,000 brand websites (415 of 1,000) | 42% ±3 |
| Food and drink brands (24 of 91) | 26% ±9 |
An FAQ = a page or section titled as one, or FAQ markup, on the pages we read. Whiskers show the margin of error.Source: remake study of 1,000 brand websites read 14–15 Oct 2026. Company list from Wikidata (CC0).
How to add FAQ schema markup
Once the answers are up, the markup is a smaller job. It’s a short block of code that tells software which line is the question and which is the answer. Bing’s guidelines (which also cover Copilot) say structured data “may support clearer grounding” (grounding is how Copilot picks the pages it answers from). A modest promise. Google stopped showing FAQ results in Search on 7 May 2026, so older advice about winning them no longer applies.
Bing adds one rule. Markup “must accurately reflect visible content.” So the block repeats your FAQ word for word, one entry per question:
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [
{
"@type": "Question",
"name": "Which countries do you deliver to?",
"acceptedAnswer": {
"@type": "Answer",
"text": "We deliver to the UK and every EU country, usually within five working days."
}
},
{
"@type": "Question",
"name": "How do I return an order?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Start a return from your account within 30 days, and we email you a free label."
}
}
]
}Your web team adds it to the FAQ page’s HTML inside this tag:
<script type="application/ld+json">
Once it’s live, give the check below your FAQ page’s address. It shows whether every question has an answer.
What else can a brand website do for AI search?
- Put a sentence or two under your headline, as Benevity does. Only 237 of the 951 brand homepages we could score have one (25%), so it’s a quick way to stand out. How to make your homepage easy for AI to quote
- Link to your Wikipedia article from your website’s code, as WIRmachenDRUCK does. It’s a small club. Just 50 of the 925 brands with a Wikipedia article have done it. How to link your website to your Wikipedia article
- Give AI search crawlers their own entry in your robots file (the short note that tells visiting software what it may read). Already, 96% of the 1,124 brand websites that answered when we asked for the file let ChatGPT and Claude search in, and few name them. A line of their own keeps them in on the day someone tidies up the rules for everyone else. How brand websites treat AI crawlers
So, back to Monday and that screenshot. Here’s a line for the client (or the boss). “Before we argue about what AI says about you, let’s make sure it can read what you say about yourself.” Then write down the five questions your buyers ask before they buy. Answer them on a page anyone can open. It makes a better Monday reply than a new brand story.
How we measured
The websites. We drew 1,477 companies at random from Wikidata, across 11 industries in eight countries (the US, Canada, Australia, the UK, Germany, France, Italy and Spain). We removed 128 by hand (public bodies, for example). Of the 1,349 left, 1,287 answered. Among them, 205 (16%) turned away an automated visitor that named itself, 37 stalled, and 45 gave errors or asked in their robots file not to be read. We read the other 1,000, each from its homepage and the pages it links to. Chain brands were read as a separate group and aren’t in these figures.
What we checked. We read each site once on 14 and 15 October 2026, keeping to its robots file. We used our website check and our free answer readiness and identity checks. We read the pages quoted here on 14 October.
Margins and limits. Shares are of the 1,000 unless another base is named. The 96% robots-file figure covers the 1,124 sites that answered when we asked for the file (with a file, no file or an error). It’s more than the 1,000 because we asked every one of the 1,287 live sites, and some that turned our page visit away still answered. The margin of error is up to ±3 points (±10 within one industry). Sites that turned automated visitors away may differ from the rest. Our draw also favours companies with a Wikipedia article. Our checks show whether a site has these basics in place, not what they change in AI answers.
Company list from Wikidata (CC0); chain brands from the Name Suggestion Index (BSD-3).
More from the study
NextPart 2 · 7 min read
Robots.txt and AI crawlers: 96% of 1,124 brand websites let ChatGPT and Claude search inPart 3 · 5 min read
Homepages and AI search: only 25% put a sentence under the headlinePart 4 · 5 min read
Only 50 of 925 brands link to their Wikipedia article in their website’s schema
See what AI says about your brand
See how well AI can read your site, then what ChatGPT says about you.