In classic search results, you can see fairly clearly who ranks ahead of you. In AI answers, this is harder. A competitor may be named far more often in ChatGPT or Perplexity, even though your website ranks solidly in Google. Or your brand is mentioned, but with a critical Sentiment describes the tone in which a brand is portrayed in AI answers, for example positive, neutral, mixed, negative or uncertain. It measures the quality of a mention, not just its presence.Open glossary entry →, outdated information or incorrect reservations.

This article goes one step beyond the question “Is my brand mentioned?”. It covers the quality of the mention:

  • How high is your Share of Voice in AI answers measures the proportion of mentions a brand receives within a prompt set compared to competitors. It indicates the competitive position in AI answers.Open glossary entry → in artificial intelligence (AI) answers?
  • How is your brand described?
  • Which sources shape the picture of your brand?
  • What do you do when your brand appears negatively?

It also explains the RankScan measure “Negative associations in AI answers” in the category 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 →.

AI Reputation describes how a brand is perceived in AI answers, that is how often, how prominently, in what tone and based on which sources it appears. It results from the brand's own website, third-party sources and model behavior.Open glossary entry → cannot be controlled directly. It is the result of your own website, third-party sources, reviews, search results, Crawling is the process by which a search engine bot fetches a URL and downloads its content. It is the first step before a page can be indexed.Open glossary entry → access, prompt context and model behavior.

Good AI monitoring makes these signals visible and helps you put them in order.

Role in the RankScan cluster: Whether your brand is mentioned or linked at all is covered in the article on AI mentions and citations. If your brand is missing from most answers on a topic, the article Page for a topic helps. This article covers share of voice, tone and the sources behind the picture of your brand.

Key takeaways are the most important insights of an article presented as a short, easily scannable list. They help users grasp central points quickly.Open glossary entry →
  • AI monitoring measures how brands appear in AI answers.
  • Share of voice in RankScan means: your share of all mentions of your brand and your competitors.
  • AI sentiment assesses the tone in which a brand is presented, for example positive, neutral or negative. RankScan itself only marks an answer as negative or not negative.
  • “Negative associations in AI answers” means: AI answers on a topic portray your brand negatively or critically.
  • Rating in RankScan: high severity · high effort · work type Editorial.
  • Negative AI sentiment often comes from old reviews, third-party sites, forums, support problems, outdated product information or unclear positioning.
  • A single ChatGPT answer is not a reliable measurement. You need two separate, repeatable prompt sets: brand-free provider questions for visibility (discovery) and branded questions for reputation.
  • RankScan measures each chosen AI assistant separately: ChatGPT, Gemini, Claude, Perplexity or Microsoft Copilot.
  • Source analysis is essential: you need to know where AI assistants get their picture of your brand.
  • Against negative sentiment, real fixes, fact-based counterevidence and consistent communication help.
  • Manipulation, fake reviews and artificial mentions are risky and counterproductive in the long term.
  • How RankScan checks it: Mark the measure as done. RankScan checks it with the next AI measurement.

When Competitors Shape the AI Answers #

RankScan used to have the measure “Competitors shape the AI answers”. It has been replaced. Today: if your brand is missing from most AI answers on a topic, RankScan opens the measure “Page for AI answers missing”. If a competitor appears in at least half of the answers, it suggests a factual comparison page. Otherwise it sharpens your service page. The article Page for a topic explains how to proceed.

Other levers have their own articles:

What Does “Negative Associations in AI Answers” Mean? #

The RankScan measure “Negative associations in AI answers” means that your brand is presented critically or negatively in AI answers.

Rating in RankScan: high severity · high effort · work type Editorial.

Examples:

“Provider X is well known, but is often criticized for high prices.”

“Users report slow support and unclear contract terms.”

“The solution appears less suitable for smaller companies.”

Negative sentiment can be accurate, outdated, exaggerated or wrong. That is why the source matters as much as the sentiment itself:

  • Which information supports the statement?
  • Is it current?
  • Does it come from your website, a review, Reddit, a ratings platform or an old article?
  • Does the statement repeat across several platforms?
  • Does it point to a real product or service issue?

AI sentiment is not just reputation cosmetics. If a critical statement appears in an AI answer, it can influence buying decisions before a user ever visits your website.

Why AI Monitoring Is Necessary #

AI answers are variable. They can differ depending on platform, prompt, language, market, time, model version, web access and available sources.

The original paper on Generative Engine Optimization (Generative Engine Optimization (GEO) covers measures that improve the chances of a brand or website being found, understood, extracted, classified and used as a source by generative AI answer systems. GEO extends classic SEO rather than replacing it.Open glossary entry →) describes generative search systems as systems that combine information from multiple sources. Website operators therefore have less direct control over when and how their content appears. That is exactly why new metrics beyond classic rankings are needed.
Source: arXiv: GEO: Generative Engine Optimization

A single manual test is therefore too weak.

Good AI brand monitoring needs:

  • fixed prompt sets,
  • repeated measurement,
  • platform comparison,
  • competitor comparison,
  • source analysis,
  • sentiment evaluation,
  • development over time,
  • documentation of changes.

A current research paper on measuring AI visibility argues that AI visibility is not one fixed value, but a measurement with uncertainty. Repeated queries say more than individual prompts.
Source: arXiv: Quantifying Uncertainty in AI Visibility

The Most Important KPIs in AI Monitoring #

1. AI Share of Voice #

Share of voice in AI answers measures what share of mentions your brand receives compared with your competitors. RankScan calculates it like this:

Share of voice = your mentions ÷ mentions of the competitive field × 100

The competitive field includes:

  • your own brand,
  • the competitors you have added,
  • the other brands named in the answers.

It does not include brands you have removed from the comparison. Depending on the setting, RankScan also leaves out brands from adjacent or unrelated markets and general terms. There is no remainder called “Other”. RankScan only counts answers to prompts without your brand name. Follow-up questions and refusals by the AI do not count as answers.

Example with made-up numbers:

Mentions in the competitive field: 96

Competitor A: 48 mentions = 50 %
Competitor B: 30 mentions = 31 %
Your brand: 18 mentions = 19 %

Where you find the value: The app does not show share of voice as a separate number. The section “Comparison with your competitors” in AI Visibility shows, for each brand, in how many answers it is named, named early and has a link to its website, plus its points. Share of voice itself is available in RankScan’s Looker Studio connector and MCP access.

What matters is not only frequency, but also prominence.

2. Answer Position #

Answer position shows how early your brand is mentioned. RankScan counts a mention as named early if it appears in the first three sentences of the answer.

In general, position can be classified like this:

PositionMeaning
First recommendationvery strong prominence
Among the first mentionsstrong visibility
Middle of the answerpresent, but not leading
Side notelow prominence
Not mentionedno AI visibility for this prompt

An early mention is more valuable than a late side note. If your brand regularly appears only in the second half of the answers, RankScan opens the measure “Brand appears late in AI answers”. The article AI mentions and citations explains it.

3. Sentiment #

Sentiment describes the tone of the mention. For your own evaluation, a general five-level scale helps:

SentimentExample
Positive“particularly suitable for Swiss small and medium-sized enterprises (SMEs)”
Neutral“is a provider of marketing automation”
Mixed“strong feature set, but complex to implement”
Negative“is often criticized for poor support”
Uncertain“information about the solution is limited”

This scale is a general method, not something RankScan displays. RankScan only marks an answer as negative or not negative.

For RankScan, recurring patterns are especially important. Individual negative phrases matter less than stable negative statements across many prompts.

4. Citation Rate #

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 → measures how often an AI answer links to your website when it names your brand. RankScan calculates it like this:

Citation rate = answers with mention and link ÷ answers with mention × 100

If the answers rarely link to your website, RankScan opens the measure “AI answers rarely link to your website”. The article AI mentions and citations explains this one too.

Perplexity is particularly helpful for source analysis because the product connects answers with visible source links. Perplexity documents its own crawlers, including PerplexityBot and Perplexity-User.
Source: Perplexity Docs: Perplexity Crawlers

5. Source Dominance #

The brand is not the only thing that matters. The domains the AI uses as sources are also decisive. RankScan shows the cited sources of the answers. The following questions are general questions for your own evaluation:

  • Is your own website cited?
  • Are competitor websites cited?
  • Are third-party sites cited?
  • Does negative sentiment come from review platforms?
  • Do statements come from old articles?
  • Are forums or Reddit threads used?
  • Are the sources current and correct?

Building a Prompt Set for AI Reputation #

Good AI monitoring starts with fixed prompt sets. Keep two sets separate:

  • Discovery set: brand-free provider questions. They show whether AI assistants name your brand on their own. In RankScan, only these answers count towards the AI The visibility index is an SEO metric that condenses many of a domain's organic Google rankings into a single value. It shows whether search visibility is rising, stagnating or falling.Open glossary entry → and share of voice.
  • Reputation set: branded questions, such as comparison or experience questions. They show how AI assistants describe your brand. They do not measure whether your brand is found.

Do not merge the two sets in your evaluation. The following prompt types are general expert knowledge.

Prompt Types #

Prompt typeExamplePurposeSet
Category prompt“Which tools help with AI visibility?”Measure share of voiceDiscovery
Comparison prompt“RankScan vs. competitor X”Check differentiationReputation
Reputation prompt“What are people’s experiences with brand X?”Check sentimentReputation
Purchase-related prompt“Which solution is suitable for Swiss SMEs?”Check decision-stage visibilityDiscovery
Problem prompt“How can I measure AI visibility?”Check topic authorityDiscovery
Competitor prompt“Which alternatives to competitor X are there?”Check conquest potentialDiscovery, as long as your brand is not in the prompt
Local prompt“Which providers in Switzerland help with Search Engine Optimization (SEO) covers all measures that help a website become more visible in unpaid (organic) search results.Open glossary entry → monitoring?”Check market relevanceDiscovery

How RankScan counts: A prompt with your brand name does not count towards share of voice and score. The brand almost always wins it. RankScan shows such prompts separately as brand awareness.

Minimum Setup #

To start, you need:

  • a fixed set of prompts without your brand name for each topic,
  • a separate set with your brand name for reputation,
  • your direct competitors,
  • several AI assistants,
  • regular repetition,
  • fixed language and region,
  • identical prompt wording,
  • documented answers.

For every measurement, document the region, language, the specific model, web access and the number of repetitions. Without these details, two measurements cannot be compared.

In a critical reputation situation, closer monitoring is worthwhile.

View Platforms Separately #

ChatGPT, Perplexity, 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 →, Gemini and Claude work differently. Do not lump the results together.

How RankScan measures: RankScan asks the prompts to 2 to 4 AI assistants, depending on your plan. You choose from ChatGPT, Gemini, Claude, Perplexity and Copilot. AI Visibility shows the results per assistant. RankScan measures Google AI Overviews separately through Google Search. This number does not count towards the score.

ChatGPT #

Depending on mode, ChatGPT can answer with or without web access. Source links are not always available. OpenAI documents different crawlers and user agents such as GPTBot, OAI-SearchBot and ChatGPT-User, each with different purposes.
Source: OpenAI Platform: Overview of OpenAI Crawlers

Perplexity #

Perplexity is strongly source-oriented and therefore useful for source analysis, citation rate and sentiment causes.

Google AI Overviews #

Google AI Overviews are embedded in Google Search. Google describes AI Overviews and AI Mode as search features in which content from the web may appear. Google does not name a special GEO technique that guarantees inclusion. The normal search fundamentals remain central.
Source: Google Search Central: AI features and your website

Important: A brand may be clearly visible in Perplexity and missing in ChatGPT, or the other way around. That is why platform segments are needed.

The access route matters too. A model queried through an An application programming interface (API) is an interface through which software systems communicate and exchange data. It defines how requests are made and answered.Open glossary entry → can answer differently from the end-user product in the browser or app. Model version, settings and web access can differ. Only treat such results as equal with this caveat.

Why Competitors Have More Share of Voice #

The following causes are general expert knowledge. Which measure RankScan derives from them is described above in the section “When Competitors Shape the AI Answers”.

  • More third-party sources: AI assistants do not rely only on a brand’s own website. Review platforms, directories, trade media and comparison articles can be decisive.
  • Clearer positioning: Competitors are often named when their category is unambiguous, such as “CRM for Swiss SMEs”. Vague self-descriptions such as “innovative digital solution” help little.
  • More Citability describes how well individual sections of a piece of content can be extracted, summarized and used as an answer by search engines and AI systems.Open glossary entry →: A research paper on AI citations found that influential sources are often more structured and contain more extractable information: definitions, numbers, comparisons and process steps.
    Source: arXiv: From Citation Selection to Citation Absorption
  • Better matching content: A study on competitive GEO found in a controlled test setup that topical relevance and position in context were the strongest drivers of first citations. Explicit prices and current timestamps also helped.
    Source: arXiv: What Gets Cited: Competitive GEO in AI Answer Engines

Why Negative AI Sentiment Emerges #

Negative sentiment rarely appears by chance.

Typical causes:

1. Real Product or Service Problems #

If many users mention the same criticism, AI systems may pick it up.

Examples:

  • poor support,
  • high prices,
  • missing features,
  • long contract periods,
  • poor documentation,
  • data protection concerns,
  • poor app reviews.

2. Outdated Information #

An old review can still shape perception, even if the product has since improved.

3. Inconsistent Communication #

When different sources provide different information, uncertainty emerges.

Examples:

  • different prices,
  • different feature descriptions,
  • contradictory target groups,
  • old product names,
  • multiple domains with inconsistent statements.

4. Third-Party Sources Dominate Perception #

If your own website does not provide clear counterinformation, review platforms or forums can shape the overall picture.

5. Missing Trust Signals #

Without sources, authors, contact details, case studies, reviews and transparent company data, a brand appears less trustworthy.

What to Do About Negative AI Sentiment #

Step 1: Isolate the Statement #

Record exactly:

Which negative statement appears?
For which prompt?
On which platform?
How often does it repeat?
Which source is cited?

Step 2: Check the Source #

Ask:

  • Is the source credible?
  • Is the criticism current?
  • Is the criticism justified?
  • Are there several sources?
  • Is outdated information responsible?
  • Is the statement wrong, or just unfavorably worded?

Step 3: Fix the Cause #

If the criticism is justified, content alone is not enough.

Examples:

  • improve support processes,
  • make pricing more transparent,
  • add data protection documentation,
  • update product documentation,
  • correct old help-center articles,
  • actively collect customer feedback.

Step 4: Create Fact-Based Counterevidence #

Create clear, crawlable content:

  • FAQ on the critical topic,
  • Accountability means it is clear which person or organisation is responsible for a piece of content and a website. It is a core trust signal.Open glossary entry → page,
  • methodology page,
  • security or data protection page,
  • product comparison,
  • updated documentation,
  • customer cases,
  • review overview,
  • changelog.

Step 5: Correct External Signals #

Check and update:

  • review profiles,
  • industry directories,
  • partner pages,
  • old guest articles,
  • app store descriptions,
  • comparison platforms,
  • social profiles,
  • knowledge panels,
  • public relations (PR) articles.

Step 6: Measure Development #

Sentiment does not change immediately. Measure it over several weeks or months.

Step 7: Report the Measure #

Mark the measure as done. How RankScan checks it: RankScan checks the measure with the next AI measurement. Once the finding is gone, RankScan reports “Fixed”. Otherwise the measure stays reported.

Avoid Manipulation #

AI reputation cannot be improved sustainably with fake signals.

Risky tactics include:

  • purchased fake reviews,
  • forum spam,
  • artificial Reddit comments,
  • A keyword is the search term users enter into a search engine to find information, products or providers. Keywords are the basis against which rankings and visibility are measured.Open glossary entry → stuffing,
  • hidden text,
  • invented awards,
  • fake authors,
  • manipulated Wikipedia entries,
  • mass-generated PR copy with no substance.

Google describes several manipulative practices in its spam policies that are intended to deceive search systems. AI systems follow their own logic. Still, the same basic rule applies: manipulative signals are risky in the long term.
Source: Google Search Central: Spam Policies for Google Web Search

Better:

  • real reviews,
  • real cases,
  • real data,
  • transparent sources,
  • clear corrections,
  • expert contributions,
  • consistent profiles,
  • visible improvements.

General Checklist: What an AI Reputation Check Can Look At #

This list is general expert knowledge. It describes a comprehensive check, not RankScan’s measurement. RankScan measures mentions, early mentions, links and cited sources per assistant, as well as negative portrayals of your brand.

A comprehensive check looks at:

  • share of voice per prompt set,
  • share of voice per platform,
  • answer position,
  • competitors before and after the brand,
  • sentiment per mention,
  • recurring points of criticism,
  • source links,
  • linked domains,
  • linked URLs,
  • Source quality shows whether important statements are backed by reliable evidence. It does not mean adding links indiscriminately, but providing robust support for verifiable claims.Open glossary entry →,
  • how current the sources are,
  • own page or third-party page,
  • wrong or outdated statements,
  • brand context,
  • prompt type,
  • language and market,
  • change over time,
  • connection with website changes,
  • crawler access,
  • external reputation sources.

Example: Negative Sentiment Due to Outdated Sources #

Situation #

A software brand is repeatedly described in AI answers as “expensive and complex”.

RankScan reports:

“Negative associations in AI answers”

Analysis #

Source analysis shows:

  • an old review from 2023 is picked up repeatedly,
  • pricing has since changed,
  • onboarding has been simplified,
  • the company’s own website does not explain the changes,
  • current customer reviews are scarce.

Solution #

  1. Update the pricing page and explain it transparently.
  2. Publish a changelog or product update page.
  3. Add FAQ content on onboarding and costs.
  4. Collect current customer testimonials.
  5. Update review platforms.
  6. Correct old external profiles where possible.
  7. Continue monitoring for 3 months.
  8. Mark the measure as done.

Result #

Negative sentiment can weaken when more current, consistent and trustworthy information becomes available.

Common Mistakes in AI Reputation Management #

Mistake 1: Testing Only Your Own Brand #

Without competitor comparison, there is no share of voice.

Mistake 2: Overvaluing Individual Answers #

A single negative result is not yet a pattern.

Mistake 3: Ignoring Sources #

Without source analysis, it remains unclear where the AI gets its assessment from.

Mistake 4: Fighting Negative Sentiment With PR Phrases #

AI systems are more likely to adopt concrete, substantiated information than advertising language.

Mistake 5: Neglecting Third-Party Sites #

Many AI answers rely on external sources. Your own website alone is often not enough.

Mistake 6: Not Checking Technical Blocks #

If 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 → or search crawlers are blocked, current information may be missing.

Mistake 7: Treating Reputation Problems Only as Search Engine Optimization (SEO) #

Real criticism must be addressed in product, service or support.

General Order: What to Do First? #

The table is a general assessment, not a RankScan rating.

ProblemUrgencyWhy
a negative statement repeats across several platformshighreputation risk
wrong or outdated information is citedhighcorrectable and business-critical
your own website is never used as a sourcehighweak source authority
competitors shape purchase-related promptshighdirect relevance to decisions
brand is named late, but neutrallymediumimprove prominence
neutral mention without a linkmediumcitation rate topic
single negative result without repetitionlow to mediumobserve

The most important rule:

Measure first, then understand sources, then fix real causes and build signals consistently.

Checklist: Check AI Reputation and Share of Voice #

Use this checklist:

  • Is there a fixed prompt set without your brand name?
  • Do you keep branded questions in a separate set for reputation?
  • Have you added your direct competitors?
  • Do you evaluate platforms separately?
  • How high is your share of voice?
  • Which competitors appear more often?
  • Who is named first?
  • What is your brand’s sentiment?
  • Do negative statements repeat?
  • Which sources are linked?
  • Are these sources current?
  • Is your own website cited?
  • Is there wrong or outdated information?
  • Are important AI crawlers allowed?
  • Is your positioning clear?
  • Are there external Authority signals are indicators of the reputation and recognised competence of a person or organisation. They correspond to the A (Authoritativeness) in E-E-A-T.Open glossary entry →?
  • Are there real customer reviews?
  • Do you measure development over time?
  • Have you marked the measure as done after implementing it?

In addition, AI readiness, Entity SEO and E-E-A-T help narrow down the cause and put the next steps in order.

FAQ About AI Monitoring, Sentiment and Share of Voice #

What is AI monitoring?

AI monitoring measures how your brand, website or product appears in AI answers. This includes mentions, position, sentiment, source links and comparison with competitors.

What does share of voice in AI answers mean?

Share of voice shows what share of all mentions in the competitive field goes to your brand. The field includes your brand, your added competitors and other providers in your market. RankScan only counts prompts without your brand name.

What should I do if a competitor shapes the AI answers?

If your brand is missing from most answers on a topic, RankScan opens the measure “Page for AI answers missing”. It asks for a comparison page or a sharper service page. The article Page for a topic explains how to proceed.

What does “Negative associations in AI answers” mean?

The measure means that your brand is presented negatively, critically or with recurring reservations in AI answers. It has high severity, high effort and work type Editorial.

How do I measure AI sentiment?

You analyze repeated AI answers for positive, neutral, mixed or negative tone and document the underlying sources.

What can I do about negative AI sentiment?

Identify the source, fix real causes, publish current counterevidence, update third-party profiles and collect genuine customer reviews. Then mark the measure as done. RankScan checks it with the next AI measurement.

Can I directly change AI sentiment?

No. You can improve the information available, but you cannot directly control AI answers. Changes need consistent signals over time.

Why is Perplexity helpful for AI monitoring?

Perplexity often shows visible source links. This helps you see which pages influence the answer.

Is classic SEO enough for AI reputation?

No. Classic SEO is the foundation. AI reputation also needs prompt monitoring, source analysis, sentiment evaluation, An entity is a uniquely identifiable thing or concept, such as an organisation, person, brand, product, place or technical term. Search engines and AI systems process information as entities and their relationships.Open glossary entry → signals and reputation beyond your own website.

Conclusion: AI Reputation Emerges From Visibility, Context and Trust #

AI reputation is more than the question of whether a brand is mentioned. What matters is how often it appears compared with competitors, how early it is mentioned, how it is described and which sources support this portrayal.

Share of voice shows your share of the competitive field. The RankScan measure “Negative associations in AI answers” shows when AI answers portray your brand critically or wrongly.

The best approach is:

  1. define a fixed prompt set without your brand name and keep branded questions separate,
  2. add your competitors,
  3. evaluate platforms separately,
  4. record share of voice, position, sentiment and sources,
  5. analyze causes instead of symptoms,
  6. strengthen content, entity and third-party sources,
  7. fix real reputation problems,
  8. mark the measure as done and wait for the next AI measurement.

This turns AI monitoring into an early-warning system for brand perception, competitive pressure and modern Search visibility describes how present and findable a website is in the organic search results. It is not the same as traffic, but reflects theoretical findability.Open glossary entry →.

Sources and Further Reading #