AI Answer Inclusion Rate: Measuring AI Visibility

Why the AI Answer Inclusion Rate is becoming a must-have metric

Search behavior is shifting. Instead of ten blue links, users increasingly read a finished AI answer. In Germany, 67% of people aged 16 and over already use generative AI at least occasionally, up from 40% the year before. According to Bitkom, the most widely used systems are ChatGPT at 43%, Microsoft Copilot at 39%, and Google Gemini at 28%.

Classic search is changing at the same time. According to Semrush, the presence of Google AI Overviews rose from 6.49% of keywords in January 2025 to 24.61% in July 2025 and fell to 15.69% by November 2025 (based on more than 10 million keywords). These AI answers cost clicks: according to an Ahrefs analysis, an AI Overview correlates with a 58% lower average click-through rate. The CTR for position 1 fell from 0.073 in December 2023 to 0.016 in December 2025.

If you don't appear in the AI answer, you lose visibility, even with a good ranking. This is exactly where the AI Answer Inclusion Rate comes in: it makes it measurable whether a brand shows up in AI answers at all.

Definition and distinction: inclusion rate, citation rate, mention rate

The AI Answer Inclusion Rate measures the share of relevant AI answers in which a brand is mentioned or cited as a source. It is the central GEO metric (generative engine optimization) for assessing AI visibility.

In practice, four terms are often mixed up. Separating them cleanly is the basis for reliable reporting:

• Inclusion rate: the umbrella term. The share of prompts for which the brand appears in the answer in any form (mention or citation). Formula: answers that include the brand divided by the total number of answers.
• Citation rate: the narrower term. The share of answers in which the domain is explicitly cited as a linked source. Relevant for platforms that show sources, such as Perplexity or Google AI Overviews.
• Mention rate: the share of answers in which the brand name appears in the body text, even without a source link. Important for systems like ChatGPT that often name brands but don't link to them.
• Share of voice (SOV): the relative value. Your own brand's share of all brand mentions or citations within a topic area. SOV puts your own inclusion rate in relation to the competition.

Rule of thumb: inclusion rate answers "Do I appear?", citation rate "Am I linked?", mention rate "Am I named?", and share of voice "How do I stack up against the competition?" In management reporting, always state which variant you are measuring.

Measurement methodology: prompt set, sampling, and LLM non-determinism

The biggest source of error in measurement is the non-determinism of language models. The same question returns a slightly different answer on every run, often with different sources.

Then there is drift over time. According to a Profound study, for the same prompts one month later (mid-July compared with mid-June 2025), 40.5% of the cited domains on Perplexity and 59.3% on Google AI Overviews were new. A single measurement is therefore not meaningful.

This leads to the sampling rule: Maximus Labs recommends querying each prompt at least 30 times per platform and measurement period. Even then, the value of a single prompt still fluctuates considerably. Reliable trends only emerge when you evaluate the entire prompt set. A practical setup for a reliable prompt set:

• Define the prompt set: 20 to 50 real user questions per topic cluster, phrased like real search queries, not like keywords. B2B DACH example: "Welche Anbieter für Lohnverrechnung gibt es in Österreich?" (What payroll providers are there in Austria?)
• Cover the intent mix: mix informational, commercial, and comparative questions, because AI systems cite differently depending on intent.
• At least 30 runs per prompt: per platform and ideally at different times, to absorb drift.
• Account for localization: use German-language prompts and set the engine's location and language to DACH, since sources and brands differ greatly by region.
• Log consistently: record the mention, citation, and position of the brand for each answer, so that mention rate and citation rate can still be evaluated separately.

If you measure without sufficient sampling, you report noise instead of trends. This is the most common methodological mistake in early GEO reports.

Multi-engine tracking: measure ChatGPT, Perplexity, Google AIO, and Gemini separately

There is no single AI visibility. Each engine has its own sourcing logic, so the inclusion rate must be measured separately for each platform.

• ChatGPT: often names brands in the text but links less often. Mention rate is the dominant metric here.
• Perplexity: relies heavily on citations with visible sources. Citation rate is meaningful and easy to measure here.
• Google AI Overviews and AI Mode: closely tied to organic search. This engine is especially relevant for DACH B2B because it builds on existing rankings.
• Google Gemini: a standalone assistant, used by 28% of people aged 16 and over in Germany, according to Bitkom.

An aggregated overall inclusion rate across all engines hides more than it shows. Report per platform and weight the results by how your target audience is distributed across platforms.

What drives AI inclusion: findings from GEO research

The question "What gets me into AI answers?" has been partly answered empirically. The Princeton GEO study tested optimizations on a benchmark of 10,000 queries from 25 domains and showed that visibility in generative answers can be increased by up to 40%. According to the study, the most effective levers were Cite Sources, Quotation Addition, and Statistics Addition, with a relative improvement of 30 to 40% in position-adjusted word count.

In practice, this means documented statistics, verbatim quotes, and clearly attributed sources increase the likelihood of being cited. Keyword optimization alone is not enough.

A second, often underestimated factor is the brand itself. According to The Digital Bloom's 2025 AI Citation Report, brand search volume, with a correlation coefficient of 0.334, is the strongest predictor of AI visibility, stronger than backlinks. A well-known brand gets mentioned more often. GEO and brand building therefore belong together.

Important for DACH B2B: classic SEO remains the foundation. According to feller.systems, 52 to 60% of AI Overview citations come from the organic top 10, and according to Ahrefs data as much as 76%. If you don't rank organically, you will rarely be cited in AI Overviews.

Linking with GSC, GA4, and conversion tracking

The inclusion rate is a visibility metric, not a business value. Only linking it with existing KPIs makes it reportable.

• Google Search Console: AI Overviews are counted in impressions, but the clicks are missing. Falling CTR with stable impressions is a sign that AI is absorbing clicks. The 58% lower CTR for AIO keywords makes this effect tangible.
• GA4 and referral attribution: capture referral traffic from AI sources such as Perplexity or ChatGPT in GA4 as a separate channel group or with hostname filters. This lets you separate AI-driven traffic from classic organic traffic.
• Conversion tracking: link AI referral sessions with conversion events to show the actual business contribution of AI visibility.

Building this bridge is essential, because management asks about traffic, leads, and ROI, not about inclusion rates.

Tool landscape for DACH and Austria

Measurement can be partly automated. Relevant for the DACH region:

• Ahrefs Brand Radar: tracks brand mentions and citations in AI answers and connects them with Ahrefs' established organic dataset.
• Semrush AI Visibility: measures presence in AI Overviews and AI engines and connects well with existing Semrush ranking data.
• Sistrix: established in the DACH market for visibility and increasingly offers AI Overview analysis; useful as a bridge to classic visibility.
• Specialized GEO monitoring tools: providers such as Profound focus on multi-engine prompt tracking with sufficient sampling.
• Manual tracking: useful as a starting point or for validation. Maximus Labs names 5 runs per prompt and platform as the minimum for directional data and 30 runs as the standard. A spreadsheet with 5 runs per prompt and platform therefore shows tendencies. Reliable values already require 1,200 queries per measurement period with just 10 prompts and four engines, which in practice is only feasible with automation.

Note for Austria and DACH: always check whether the tool supports German-language prompts and regional engine settings. Otherwise, international tools measure against the US market.

Benchmarks and realistic targets in a B2B DACH context

Read absolute benchmarks with caution, as they depend heavily on industry, prompt set, and engine. As a pragmatic guide, the following ranges have become common: an inclusion rate below 5% counts as practically invisible, 15 to 30% as a strong presence, and above 30% as category-leader level. These values are based on experience, not on studies, and should be calibrated internally.

What is well documented, however, is that AI visibility matters more in technical B2B markets than elsewhere: according to feller.systems, AI Overviews appear on 70% of SERPs in B2B technology and SaaS, up from 36% in May 2024. In these segments, a low inclusion rate is a direct competitive disadvantage.

Set relative targets instead of absolute ones: improve your own inclusion rate compared with the previous month and increase your share of voice against your three most important competitors.

Competitive analysis: share of voice and competitive displacement

AI answers usually name only a few providers. Visibility here is a displacement contest. Two analyses are central:

• Share of voice against competitors: record all brands named for each prompt and calculate your share. This shows whether you even belong to the AI's relevant set in your topic area.
• Competitive displacement: track over time which brands appear and which disappear. Because of the monthly citation drift of 40 to 60%, you need sufficient sampling here too; otherwise, you will mistake noise for displacement.

This view is more valuable than your own inclusion rate in isolation, because it shows whether you are gaining or losing ground.

Practical workflow and reporting

A continuous workflow connects research, optimization, and measurement:

1. Prompt and keyword research: collect real user questions for each topic cluster and turn them into a versioned prompt set.
2. Measure the baseline: capture inclusion, citation, and mention rate per engine with at least 30 runs per prompt.
3. Optimize content: add documented statistics, quotes, and clear sources, since according to the GEO study they deliver up to 40% more visibility, and secure your organic top-10 position.
4. Strengthen the brand: build brand search volume, the strongest predictor of AI visibility.
5. Measure and report continuously: compare inclusion rate, share of voice, and AI referral conversions every month in management reporting.

For reporting: show AI visibility alongside rankings, organic traffic, CTR, and conversions, not in isolation. In SEO forecasting, you can plan the expected CTR erosion from AI Overviews as a scenario, based on the documented CTR reduction. This turns a new metric into a figure that matters for management decisions.

Common mistakes

• Single measurement instead of sampling: one run per prompt ignores non-determinism and produces noise.
• Aggregating engines: one overall value across all AI systems hides platform-specific realities.
• US prompts for DACH: international defaults measure the wrong market and the wrong brands.
• Pitting GEO against SEO: AI inclusion builds on organic top-10 rankings; classic SEO remains the foundation.
• Inclusion rate as an end in itself: without a link to traffic and conversions, reporting lacks the business connection.