Most Brand Websites Let AI In. Few Give It Anything to Quote.

The shortfall sits in the basics a machine uses to identify a company and decide whose words to repeat. Two thirds of the sites (67 percent) do not list their official profiles in their structured data, and 86 percent show no author names. Nine in ten brands (92 percent) share a name with another entity or have an identity split across sources.

The study is based on about 50 brand websites that CiteSurge audited in depth between July and October 2026. The brands include businesses in hospitality, games, online gambling, law, venture capital, software, healthcare, media and events, as well as tourism boards, mostly in Europe, and none is named here. The methods section explains how each figure was measured and how the brands were anonymised.

Key takeaways

• Finding 1: crawler access is no longer the hard part. 92 percent of sites let every major AI crawler in, yet 100 percent of audits scored higher on technical access than on quotability, by 37 points on average.
• Finding 2: the identity basics are rarely complete. Only 4 percent of sites publish all four identity and authorship basics: structured data, Organization markup, sameAs links and author names. 35 percent publish none.
• Finding 3: outside sources are stronger than the brands' own markup. In 75 percent of audits, the brand scored higher on how other websites describe it than on its own structured data.
• Finding 4: expert names are missing from the articles. 86 percent of sites show no author names, and 59 percent of the sites that name experts somewhere do not credit them as authors.
• Finding 5: a name problem is the norm. 92 percent of brands share a name with another entity or have a split identity, and about two in three face a direct name clash.
• Finding 6: many homepages give AI nothing to quote. 35 percent of the homepages assessed hold no quotable answer, and 24 percent of sites need JavaScript to show their main text.
• Finding 7: comparison pages are rare. 92 percent of the sites assessed have no comparison or alternatives page of their own.

1. Getting crawled is no longer the hard part

Technical access is the highest average score in this study, and quotability sits far below it. On a 100-point scale, technical access averaged 85.5, quotability 48.6 and structured data 33.7. In no single audit was the gap between access and quotability smaller than 14 points. Only 2 percent of sites scored 70 or more out of 100 on quotability, and none reached 75.

Technical access measures whether AI crawlers can reach and read a site. Quotability measures whether its pages contain passages an AI answer can lift and use on their own. Structured data measures how completely a site publishes machine-readable markup about who the company is.

CiteSurge's overall score is below 60 out of 100 for 51 percent of sites, and only 4 percent of sites reach 75 or higher.

Our reading is that, for most of these brands, the crawler is no longer what stands between them and an AI answer. What the pages say, and how plainly the site says who the company is, now decides whether an answer uses them.

The 10 pages that matter most to your buyers need a short passage near the top that says what you do, for whom, where and at what price. Write it so it still makes sense when someone quotes it on its own.

2. The identity basics go missing together

We looked for four things that tell a machine who a company is and who speaks for it: structured data on any page, Organization markup, sameAs links to official profiles, and author names. Only 4 percent of sites had all four, and 35 percent had none. Of the rest, 16 percent had one, 24 percent had two and 20 percent had three.

Counted one at a time, 49 percent have no structured data on any page we reviewed, 55 percent have no Organization markup and 67 percent have no sameAs links. Only 31 percent have both Organization markup and sameAs links, the pair that states a company's name and points to its official profiles. The other 69 percent lack at least one of the two.

The likely result is that most of these companies leave each AI engine to work out who they are from other sources.

Your homepage needs one Organization block. Give it your legal name, logo and location, and link every official profile you already have.

3. Other websites describe these brands better than the brands do

Outside sources are stronger than the brands' own markup in most of these audits. Brand authority, our measure of how a company appears on other websites, scored higher than structured data in 75 percent of audits. The average gap was about 20 points: 53.3 against 33.7.

Public records are already in place for some of these brands: a Wikipedia article exists for 26 percent and a Wikidata entry for 40 percent. Of the brands with a Wikipedia article, 67 percent have no sameAs links on their own site. For brands with a Wikidata entry, the share is 68 percent. The companies' own sites do not point back to the records that name them.

The result is a brand that is known in the places AI engines read, with a website that does not confirm it.

Your sameAs markup should list every official profile, starting with Wikipedia and Wikidata where they exist. Then open each record and check that it names your company and not a namesake.

4. The experts are on the site, but their names are not on the articles

Author names are missing from the pages on 86 percent of sites, and only 14 percent show them. The expertise is often there. Among the sites assessed, 38 percent name experts somewhere, and 59 percent of those still do not credit them as authors.

The people are named in one place, and the advice is published without them. An article credited to "admin", or to no one, tells neither a reader nor an engine who stands behind it.

Put a byline and a two-line factual bio on every article and guide, and link each byline to a profile page for that person.

5. Nine in ten brands have a name problem

Name problems are the norm in these audits: a name clash or an identity split across sources affects 92 percent of brands. About two in three (67 percent) face a direct clash: another company, product, place or person uses the same name, or a public lookup resolves the name to the wrong entity. The rest have an identity split without a shared name, such as a parent company's record standing in for a subsidiary, or a former name that never left a profile.

The namesakes in these audits include a sports team, a town, a stock-market ticker and a defunct app.

When two entities share a name, an engine needs the brand's own site to say plainly which one it is. In 55 percent of audits, the site has no Organization markup to say it.

Open your About page with one plain sentence that gives your full legal name, what you do, where you are based and how you differ from the namesake you already know about.

6. Many homepages offer nothing to quote

A quotable answer is missing from 35 percent of the homepages assessed: no short block answers a basic buyer question on its own, such as what the company does, for whom and where. On 24 percent of sites, the main text needs JavaScript to appear, so a crawler that does not run JavaScript sees little of it.

Two optional AI text files are missing from most sites: llms.txt from 80 percent and llms-full.txt from 96 percent. The bigger gap in these audits is on the pages themselves.

A quick test needs only your browser: open your homepage's page source and search for your main headline. If it is missing, your text depends on JavaScript. If it is there, check that the first screen holds a sentence saying what you do, for whom and where.

7. Few brands publish their own comparison page

Comparison pages are rare: among the sites assessed, 92 percent have no comparison or alternatives page among the pages we reviewed. Buyers ask head-to-head questions, and a brand without its own page leaves the answer to others.

To see what that looks like, we checked search results, which show the pages an engine can draw on. These are search results, not AI answers. For one software vendor, the top result for its head-to-head against a competitor was the competitor's own comparison article. For a consumer-electronics maker, a comparison website supplied some of the results and the maker's own site appeared in none.

Our reading is that when a brand publishes no comparison, its rivals and the review sites write it instead.

One comparison page is enough to start: compare yourself with your two closest alternatives in a table that shows who each option suits, the price, the limits and the date you checked.

How we measured this

What the figures come from

Every figure in this study is drawn from about 50 of CiteSurge's own AI visibility audits, run between July 10, 2026 and October 19, 2026 on the same audit format, so every score compares like with like. Where a site was audited more than once, the latest run counts. Each audit reviews a sample of the site's pages, its crawler rules and its structured data, and reviews how the brand appears in public records and on other websites. The audits are outside-in: they use only what any visitor, crawler or search engine can see. How the 9 AI engines answer buyer questions is a separate part of each audit and sits outside this study. The methodology page describes CiteSurge's general approach to observing, diagnosing and checking AI visibility.

How the figures were computed

• A percentage is the share of audits in which a finding was present, out of the audits where that check could be assessed. A check that could not be assessed for a site, such as a homepage that could not be read, leaves that site out of the percentage and is not counted as a no. That is why some findings say "assessed".
• An average is a simple average across audits, and a gap is measured inside each audit first and then averaged.
• Percentages are rounded to whole numbers and scores to one decimal, so a group of percentages can total 99 or 101.
• The access result is not driven by a few sites. With the audits that had the thinnest evidence set aside, every remaining audit still scored higher on technical access than on quotability, and the smallest gap was still 14 points.

Terms used in the findings

• AI visibility: whether and how AI engines name, describe and cite a brand when buyers ask questions. It covers generative engine optimization (GEO), answer engine optimization (AEO) and AI agent readiness.
• Every major AI crawler allowed: the site is open to all the major AI crawlers we tested, in both its crawler rules and its server responses.
• Structured data on any page: at least one page we reviewed contains JSON-LD markup.
• Organization markup: a structured-data block that gives the company's name and basic details, as described in Google's Organization structured data documentation.
• sameAs links: entries inside that markup that connect the site to the company's official profiles, such as Wikipedia or LinkedIn, using the Schema.org sameAs property.
• Author names: articles or guides show a visible byline that names a person.
• Named experts shown: the site shows experts or staff by name somewhere, whether or not it credits them as authors.
• Name clash or split identity: the public records show another entity with the same or a near-identical name, a lookup that resolves the name to the wrong entity, or an identity split across names, records or profiles.
• Comparison or alternatives page: a page on the brand's own site that shows how the brand compares with named alternatives.
• Text files for AI: some sites use optional files named llms.txt and llms-full.txt to describe themselves to AI systems.

How the brands were anonymised

No brand, domain, person or client is named. Examples are described by type, and details that could identify a brand are generalised or left out.

What this study does and does not show

The audited brands are CiteSurge clients and prospective clients, so the figures describe these brands and not all websites. Each audit records a site on the day it was audited. Sites change, and a finding can be fixed the next week, which is why CiteSurge re-runs these checks over time instead of treating one audit as final.

The study shows what the audits found and where gaps cluster. It does not show that fixing a gap makes an engine cite a brand. A missing comparison page means none turned up among the pages we reviewed, and the two search-result examples in finding 7 sit outside every percentage. The name clash record comes from each audit's reading of public records, and the split into a direct clash and an identity split is a grouping of those records.

Frequently asked questions

Why does a name clash matter for AI answers?

A shared name can lead an AI engine to the wrong company, so the brand's own site needs to say plainly which entity it is. In these audits, 92 percent of brands have a name clash or a split identity, and 55 percent of sites have no Organization markup to settle it.

Is a missing llms.txt file the main gap?

The pages themselves are the bigger gap in these audits: 35 percent of the homepages assessed hold no quotable answer, and 24 percent of sites need JavaScript to show their main text. The llms.txt file is missing from 80 percent of sites.

What should a brand fix first?

Two fixes are a sound place to start: a short passage near the top of each key page that says what the company does, for whom and where, and one Organization block with sameAs links to every official profile. A CiteSurge audit then shows which of the remaining gaps cost the brand answers on the questions its buyers ask.

Cite this study

The study can be cited as: CiteSurge, "Most Brand Websites Let AI In. Few Give It Anything to Quote.", October 26, 2026, citesurge.com/insights/ai-visibility-audit-study-2026. For a chart or a single figure, credit it as: Source: CiteSurge AI visibility audits, 2026.

How CiteSurge gets your brand cited in ChatGPT and AI search

Seven findings, one pattern: AI can get in, and it finds too little to quote or to identify the brand by. Each gap can be fixed, and most sites have several. What a team cannot see from its own website is which of them is costing it answers today, on the questions its buyers actually ask.

CiteSurge is the AI visibility company for GEO and AEO, and it shows you exactly that, engine by engine. It puts your buyers' questions to 9 AI engines (ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews, Google AI Mode, Bing Copilot, Grok and Meta AI), records who is named and who is cited, and tracks your share of voice against the competitors you pick. It also tracks the keywords AI engines like to cite and the search demand behind each one, so you know which to go after first. An AI agent readiness test then runs a real buyer task on your own site and shows every step, down to the point where the agent gave up.

Every answer about your brand is checked as well. CiteSurge flags answers that describe you wrongly or out of date and tests whether AI recommends you for the needs your buyers state. Each gap is traced to the page or outside source behind it, and each fix comes with an estimate of its revenue potential, so the work starts where the return is largest.

CiteSurge can do the work itself: the pages, the content and the outside mentions that get your brand cited, published at the level of autopilot you choose. Enterprise teams can hand us the whole programme, and we lead the approved content and website changes.

CiteSurge connects to all the apps and CMS platforms you already use, such as Google Search Console, GA4, HubSpot, Shopify and WordPress, so you see what every change did to your AI answers and your revenue. An alert tells you when an answer about your brand moves.

AI answers are changing every day, and each change decides which brand a buyer hears about. CiteSurge is on it 24/7, following every shift across all 9 engines and acting on it before your competitors do. If you want the best AI visibility team working for you as AI takes over search, send us a message and we'll show you how to get ahead of your competition.