Classic Search Engine Optimization (SEO) covers all measures that help a website become more visible in unpaid (organic) search results.Open glossary entry → asks: Does your website rank in Google? AI Visibility describes whether and how visible a brand, website or person is in AI-generated answers. It measures mention, position, description and source linking across multiple platforms and prompts.Open glossary entry → also asks: Do AI assistants name your brand, and do they link to your website?
In ChatGPT, Perplexity, Gemini or Claude, nobody sees a list of ten results. People read a summarized answer. Your brand appears early, late or not at all. And your website is either listed as a source or missing.
This article explains two RankScan measures in the category AI Visibility:
- “Brand appears late in AI answers”: Your brand is named, but mostly only after other brands.
- “AI answers rarely link to your website”: Your brand is named, but your website is rarely listed as a source.
If your brand is missing from most answers of a topic, a different measure applies: “Page for AI answers missing”. It is covered in the article on the topic page.
AI visibility is not a placement you can buy or force. It is the likelihood of being named early and linked in many answers.
- AI visibility measures whether and how your brand appears in AI answers: named, named early, with a link.
- RankScan asks your prompts to 2 to 4 assistants, depending on your plan. You choose from ChatGPT, Gemini, Claude, Perplexity and Microsoft Copilot. Only prompts without your brand name count.
- “Brand appears late in AI answers” means: In at least three answers that name your brand, it stands on average in the back half of the answer, measured up to the last brand named.
- “AI answers rarely link to your website” means: Of the answers that name your brand, fewer than 20% link to your website.
- Rating in RankScan: “Brand appears late” has medium severity · medium effort · work type Editorial. “Rarely link” has high severity · medium effort · work type Editorial.
- You get named earlier mainly through independent sources. You get linked through a clear target page with facts as text.
- How RankScan checks it: RankScan checks both measures with the next new AI data. Once the finding is gone, RankScan reports “Fixed”.
What Does AI Visibility Mean? #
AI visibility describes how visible a brand is in the answers of AI assistants.
Typical prompts look like this:
Which SEO tools are suitable for Swiss SMEs?
Which agencies help with marketing automation in Mautic?
Which timber builder in Central Switzerland builds apartment buildings?
Three questions matter for such prompts:
- Is your brand named?
- Is it named early, before or together with the other brands?
- Is your website linked as a source?
Google describes AI Overviews are AI-generated answer boxes from Google that appear above the classic search results. They summarize information from the web and can change click distribution.Open glossary entry → and AI Mode as features that use content from the web. Websites can still appear in them. Google names no special technique that guarantees inclusion. The normal foundations of search remain central. Source: Google Search Central: AI features and your website
How RankScan Measures AI Answers #
RankScan regularly asks the AI assistants you choose the prompts of your topics: 2 to 4 of ChatGPT, Gemini, Claude, Perplexity and Copilot, depending on your plan. In every answer, RankScan looks for your brand, the brands of your market and the linked sources.
Not every answer counts for the two measures:
- Only prompts without your brand name: A prompt that already names your brand always wins it. So it does not count.
- Only real answers: A follow-up question or a refusal by the assistant is not a measurement.
- Only the latest answer: Per prompt and assistant, the most recent answer counts.
Google AI Overviews do not count in these two measures.
The answers produce three values:
| Value | Question | Measure when the value is weak |
|---|---|---|
| Named | In how many answers does your brand appear? | “Page for AI answers missing” |
| Named early | How early in the answer does your brand appear? | “Brand appears late in AI answers” |
| With link | How often does an answer that names your brand also link to your website? | “AI answers rarely link to your website” |
How foreign pages used by AI answers lead to a measure of their own is explained in the article on brand mentions on cited pages.
What Does “Brand appears late in AI answers” Mean? #
The RankScan measure “Brand appears late in AI answers” means: When your brand appears for a topic, it typically appears late in the answer.
Rating in RankScan: medium severity · medium effort · work type Editorial.
How RankScan Measures “Late” #
RankScan does not simply count the sentence in which your brand appears. A long answer would otherwise always have a late brand. So RankScan measures the share of the answer:
- RankScan finds the sentence in which your brand first appears.
- RankScan finds the sentence in which the last brand named first appears.
- The share is the first value divided by the second. 0 means: your brand is at the start. 1 means: your brand is level with the last brand named.
The measure is created when the share is above 0.5 on average. RankScan needs at least three answers that name your brand. Answers without your brand do not count here. If an answer names only your brand, your brand is at the front there.
An example: An answer names two competitors in the first sentence, another in the third and your brand in the fifth. Your brand is level with the last brand. Its share is 1.
Why Brands Appear Late #
RankScan explains the measure like this: AI answers name first the brands that independent sources list for the topic. Your own pages alone rarely change the order. Trade media, industry reports and awards weigh more.
A late mention is no verdict on your quality. It shows how the sources of the answer describe your market.
The Sources of These Answers #
The page of the measure shows the list “Sources of these answers”. It names up to five foreign pages. Each comes from an answer in which another brand stands before yours.
For each source, RankScan shows:
- in how many of these answers it is cited,
- whether it names you, or which competitors it names instead of you,
- its kind: Competitor, Portal or Neutral source.
At an equal number of answers, a source that names competitors and not you comes first. There, your brand is missing from the picture the assistant reads.
RankScan determines the kind like this:
| Kind | What it covers |
|---|---|
| Competitor | a domain of a tracked competitor |
| Portal | a comparison portal or directory, also a blog post that names at least two competitors |
| Neutral source | any other page, for example trade media, an association or a forum |
Procedure #
The steps follow the kinds in the source list. RankScan shows only the steps of the kinds that occur in your finding.
Always first:
- Read the AI answers and note which brands are named before yours.
- Name on your pages the facet in which you are stronger than these brands.
- Back it up with references: project, year, location, partner, award. Add them as structured data too.
For sources of the kind Competitor:
- You cannot change pages of competitors.
- Answer the questions of these pages on your own page, with your own details.
For sources of the kind Portal:
- Check your listing on each portal: profile, offer, images and reviews.
- If the listing is missing, clarify the terms: fee, partner programme or registration.
For neutral sources:
- Write to the operators of the neutral sources.
- Suggest a mention with a concrete occasion, for example a project, a study or an award.
Finally:
- Mark the measure as done.
How to propose a mention on a foreign page is explained in detail in the article on brand mentions on cited pages.
How RankScan Checks It #
Mark the measure as done once you have implemented it. RankScan then checks it with the next new AI data. If the share is no longer above 0.5 on average, RankScan reports “Fixed”. Otherwise the measure stays reported.
Expect weeks rather than days. An assistant only picks up a new mention once it reads the source again.
What Does “AI answers rarely link to your website” Mean? #
The RankScan measure “AI answers rarely link to your website” means: AI answers about a topic name your brand but rarely link to your website.
Rating in RankScan: high severity · medium effort · work type Editorial.
How RankScan Measures the Citation Rate #
RankScan calculates the Citation Rate measures the share of relevant AI answers in which a website is linked or named as a source. It is a monitoring metric, not an official ranking factor.Open glossary entry → like this:
Answers with a link to your website
÷
Answers that name your brand
The measure is created when the citation rate is below 20%.
An example with made-up numbers: 20 answers name your brand, 3 of them link to your website. The citation rate is 15%. The measure is created.
Answers without your brand do not count in the denominator. So the citation rate does not measure how often you appear. It measures how often your website is a source when you appear.
Why AI Answers Do Not Link to Your Website #
RankScan explains the measure like this: AI assistants link to sources that state facts clearly and as readable text. If the facts are scattered or only in a PDF, they tend to link to other websites.
Without a link, the answer still names your brand. But people do not reach your website through the answer.
Procedure #
The measure has five steps and the final step:
- Choose for each topic the page that AI answers should link to. If there is none, create one.
- Write clear, quotable statements there.
- Always name your brand, services, projects and places by their exact name.
- Put key facts like figures, dates and prices on the page as text, not only in a PDF or image.
- Have your developers or agency check that the page has a stable address and a self-referencing canonical tag.
- Publish the page and mark the measure as done.
For step 1, the page you chose as “Page for this topic” in the topic settings is usually the right one.
How RankScan Checks It #
Mark the measure as done once you have implemented it. RankScan then checks it with the next new AI data. If the citation rate is no longer below 20%, RankScan reports “Fixed”. Otherwise the measure stays reported.
Why AI Answers Fluctuate #
AI answers are not always the same. The same prompt can give a different answer depending on time, model and web search.
A research paper on measuring AI visibility advises treating metrics such as the citation rate as estimates with uncertainty. Repeated measurements say more than a single query. Source: arXiv: Quantifying Uncertainty in AI Visibility
That is why “Brand appears late” needs at least three answers that name your brand. “Rarely link” has no such lower limit: If only one answer names your brand without a link, the citation rate is 0%.
For you, this means: Do not judge a single answer you see in ChatGPT yourself. Judge the trend in RankScan.
Quotable Pages: What Helps With Links #
The five steps of the measure are the core. This Expertise is the subject-matter competence behind a piece of content. It is the second E in E-E-A-T and should be visible through author details and supporting evidence.Open glossary entry → helps you implement them.
Deliver the Answer Directly #
Start important sections with a clear statement.
Weak:
In today's digital world, there are many ways to increase visibility.
Stronger:
Muster AG builds apartment buildings in prefabricated timber element construction, with over 40 projects in Central Switzerland since 2009.
Facts as Text, Not in an Image #
Prices, dates, key figures and references belong in the text of the page. A PDF or an image can add to it but not replace it.
Clear Structure #
Use subheadings, tables, lists and short definitions. A research paper found: pages with high influence in AI answers tend to be longer, better organized and richer in definitions, figures, comparisons and process steps. The paper describes an association, not a cause. Source: arXiv: From Citation Selection to Citation Absorption
Stable Address #
Every important statement sits on a stable address with a The canonical tag is an HTML link element that tells search engines the preferred (canonical) URL of a page. It helps consolidate duplicate or similar URLs onto a main version.Open glossary entry → tag.
Avoid:
- changing addresses with parameters,
- content only behind a login or a form,
- content that only appears with JavaScript.
Getting Named Earlier: Independent Sources #
For “Brand appears late in AI answers”, your own website is rarely enough. The answer follows the sources the assistant reads.
Three ground rules:
- Never write to competitors. Answer their questions on your own page, without copying text.
- Ask with an occasion. A project, a study or an award is a reason for a mention. In our experience, a plain request rarely works.
- Stay genuine. Paid mentions without labelling, made-up reviews or fake forum posts damage trust.
If foreign pages that AI assistants use for a topic do not name your brand, RankScan opens the measure “Relevant sources without your brand”. The article on brand mentions on cited pages explains the process. If AI answers cite a competitor's page instead of yours, the measure is called “A competitor page is cited instead of yours”. It has its own article.
Technical Requirement: Assistants Must Read Your Page #
An assistant can only link to what it can reach. If The robots.txt is a publicly accessible text file in a site's root directory. It tells crawlers which URLs they may fetch and therefore controls crawling.Open glossary entry → or a firewall blocks its crawlers, the chance of a link drops.
OpenAI documents several crawlers with their own purpose, including GPTBot, OAI-SearchBot and ChatGPT-User. Website operators control them through robots.txt.
Source: OpenAI Platform: Overview of OpenAI Crawlers
Perplexity documents the crawlers PerplexityBot and Perplexity-User.
Source: Perplexity Docs: Perplexity Crawlers
If your robots.txt blocks AI crawlers are automated bots that fetch web pages for AI providers in order to find content, make it available for search functions, retrieve it for answers, or, depending on the provider, use it for model training.Open glossary entry →, RankScan reports this in a measure of its own: “AI crawlers blocked in robots.txt”. More on this in the article on controlling AI crawlers.
You do not need an llms.txt for this. RankScan lists it only as a note with no action required.
General Checklist for an AI Visibility Check #
This list is general. It shows what a comprehensive check can look at. In these two measures, RankScan measures only part of it.
RankScan measures in these measures:
- whether your brand appears in the answers to your prompts,
- how early in the answer it appears,
- whether an answer that names your brand links to your website,
- which foreign sources the answers cite and whether they name you,
- per chosen assistant: ChatGPT, Gemini, Claude, Perplexity or Copilot.
You check yourself:
- whether the answer describes your brand correctly,
- whether your positioning reads the same on the home page, the about page and your profiles,
- whether your target page shows the facts as text,
- whether trade media, associations and portals list you.
Example: Brand Is Named, but Late and Without a Link #
The example is made up.
Starting Point #
A timber builder is named in many answers to prompts about its core topic. RankScan shows two measures:
Brand appears late in AI answers
AI answers rarely link to your website
Analysis #
The source list shows an industry portal and a trade article. Both name two competitors, not the timber builder. Its service page describes the offer only in general terms. References are in a PDF.
Solution #
- The builder checks its listing on the industry portal and adds offer, images and references.
- It offers the editors of the trade article a completed project as an occasion.
- It chooses its service page as the target page of the topic.
- It puts projects with year, location and construction method on the page as text.
- Its agency checks the address and canonical tag of the page.
- It marks both measures as done.
Result #
RankScan checks both measures with the next AI data. Whether and when the answers change is up to the assistants. There is no guarantee.
Common Mistakes #
Mistake 1: Treating Google Rankings as AI Visibility #
Good rankings help. But they guarantee neither a mention nor a link in AI answers.
Mistake 2: Overrating a Single Answer #
An answer you see in ChatGPT yourself is not a measurement. Look at the values across all prompts and assistants.
Mistake 3: Asking Competitors for a Mention #
No competitor names the competition voluntarily. Answer their questions on your own page.
Mistake 4: Marketing Language Instead of Facts #
Assistants link to clear, verifiable statements. Interchangeable advertising sentences give them nothing to quote.
Mistake 5: Facts Only in a PDF #
References, prices and dates in a PDF or an image are hard to use for an answer. Put them on the page as text.
Mistake 6: Blocking Crawlers #
If you block OAI-SearchBot, PerplexityBot or other crawlers of assistants, you give up the chance of a link.
Mistake 7: Fake Mentions #
Made-up reviews, fake forum posts or spam on Wikipedia are risky and damage trust.
Checklist #
- Is a target page chosen for each topic?
- Does the target page name brand, services, projects and places by their exact name?
- Are figures, dates and prices on the page as text?
- Does the page have a stable address and a self-referencing canonical tag?
- Have you noted the brands named before yours?
- Have you named the facet in which you are stronger?
- Are your listings on the portals of the source list complete?
- Have you asked neutral sources with a concrete occasion?
- Have you refrained from writing to competitors?
- Are the crawlers of the assistants not blocked?
- Have you marked the measure as done?
In addition, the topic page, brand mentions on cited pages, entity SEO and thin content and high-quality content help you narrow down the cause and choose the next steps.
FAQ About AI Visibility, Mentions and Links #
What is AI visibility?
AI visibility measures whether and how your brand appears in the answers of AI assistants: named, named early and with a link to your website. The article on Generative Engine Optimization covers the bigger picture.
Which assistants does RankScan measure?
RankScan asks your prompts to 2 to 4 assistants, depending on your plan. You choose from ChatGPT, Gemini, Claude, Perplexity and Copilot. For the measures, only prompts without your brand name and only real answers count.
What does “Brand appears late in AI answers” mean?
Your brand appears in at least three answers, but on average in the back half of the answer. The steps follow the kind of the sources of these answers: competitor, portal or neutral source.
What does “AI answers rarely link to your website” mean?
Of the answers that name your brand, fewer than 20% link to your website. The lever is one clear target page per topic with facts as text.
What is the citation rate?
The citation rate is the share of answers with a link to your website among the answers that name your brand.
What do I do if my brand is not named at all?
Then the measure “Page for AI answers missing” applies. It is covered in the article on the topic page.
Should I ask competitors for a mention?
No. You cannot change pages of competitors. Answer their questions on your own page, with your own details.
Do I need an llms.txt?
No. RankScan lists it only as a note with no action required.
When is the measure fixed?
RankScan checks it with the next new AI data. Once the finding is gone, RankScan reports “Fixed”. Otherwise the measure stays reported.
Can I guarantee AI mentions?
No. AI answers depend on assistant, model, prompt, web search and the sources available. You can improve the conditions, but you cannot force a mention.
Conclusion: Named Early and Linked #
AI visibility has several stages. Your brand must be named, stand early and bring your website along as a source. RankScan measures each stage separately and opens a measure of its own for each.
“Brand appears late in AI answers” works through the sources of the answers: check portals, ask neutral sources with an occasion, answer competitors' questions on your own page. “AI answers rarely link to your website” works through one clear target page per topic.
The process:
- open the measure in RankScan and read the sources or the topic,
- work through the steps of the measure,
- mark the measure as done,
- wait for the next AI data.
