AI Visibility Score: What's Good? 2026 Benchmarks

A good AI visibility score for a home-services company is an appearance rate of 34.9% or higher, the top-quarter line across the 13 home-services companies Cheers tracked in the 28 days ending September 2, 2026. Leaving out questions that name the company itself, the median company was named in 29.5% of the AI answers we checked for it, and the bottom quarter sat under 12.8%. On a set of 100 buying questions, that spread is the difference between being named in fewer than 13 answers and in 35 or more, and every answer you are missing names somebody else.

Last verified: September 28, 2026.

The benchmark bands

An AI visibility score, the way I use the term, is an appearance rate: the share of tracked AI answers that name your business. If you do not know your own number yet, the AI Visibility Grader will measure one location in a few minutes. The 13 companies fell into four bands:

• Below 12.8%: bottom quarter. The business is missing from about seven tracked answers in eight.
• 12.8% to 29.5%: below the median company. Named sometimes, rarely by default.
• Around 29.5%: the median company, named in about three answers in ten.
• 34.9% or higher: top quarter. This is the line I call good.

The mean across companies was 25.2%, a little under the median because a few companies near the bottom pull it down. The median company was cited as a source in 20.2% of its answers, and every one of the 13 companies was named at least once, so total absence is rare; frequency is the problem. There is no top-10% band. With 13 companies a 90th percentile would sit between the two highest companies, so we do not publish one.

Those rates leave out branded prompts, questions such as "would you hire [your company] for a sewer backup?" that put the business name in the question and get it back almost every time. Three of the 13 companies track them. Left in, they would lift the mean to 27.9% and the bottom-quarter line to 22.7%, which measures whether an engine repeats a name you typed, not whether it recommends you. If your own tracking includes questions with your name in them, take them out before you compare against these bands.

Read the scope before you borrow the numbers. Each company is measured once across all of its tracked locations, prompts, and engines: 183 locations and 958,754 AI answers across 593 tracked prompts between them. The panel is home services only, so a dentist or a law firm should not read these bands as their own, and 13 companies is a small panel. Treat the bands as a working benchmark, not a law of nature. The full cuts are in the Home Services AI Visibility Index.

One company, four different scores

The same 13 companies score very differently depending on which engine you ask. Median company appearance by engine, same window: Google AI Mode 36.0%, ChatGPT 24.1%, Gemini 21.3%, Perplexity 13.9%. That is 22.1 points between the best and worst engine for the same businesses.

Citation rates split differently again. The company mean citation rate was 31.0% on Google AI Mode, 23.4% on Perplexity, 11.5% on ChatGPT, and 10.8% on Gemini, so a company can be named often on one engine and almost never used as a source on another. ChatGPT is also the one to watch on position. Across the 10 companies with ranked ChatGPT answers, the median company's average position when named was about 2.7. Third on a four-name list is a different commercial outcome from first, and an appearance rate on its own will not show you that.

The wording of the questions moves the number too, which is why a score needs its question list attached. Pooled across every check, questions asking for the best or top companies without a place name returned the tracked company 28.4% of the time; our panel is still too small to publish most other question shapes, including near-me questions, which only nine companies track. If a vendor quotes you one score, ask which engines and which questions are behind it, and whether any of those questions contain your own name, before you compare it with anything. The source comparison explains why one engine cannot stand in for the rest.

An AI visibility score is not an SEO visibility score

Search for "visibility score" and most of what comes back describes an SEO metric. Semrush says its Visibility % "is based on the click-through rate (CTR) that shows a website's progress in Google's top 100 for keywords" in a tracking campaign. SISTRIX weights each ranking "according to the search volume of the keyword and the expected click-through-rate (CTR) on the measured, organic position." Both are built from where your pages rank in classic Google results.

An AI visibility score starts somewhere else. It counts whether an answer from ChatGPT, Gemini, Perplexity, or Google AI Mode named the business at all, and it is read from saved answers you can audit line by line. The two can move in opposite directions. A service page can sit on page one for "water heater replacement" while the AI answer to "who should I call to replace my water heater" names three competitors and a directory.

Keep both, on separate lines of the report. Never average an SEO visibility percentage into an AI appearance rate; they share a name and nothing else.

Use appearance rate before a blended score

AI visibility tools calculate their scores in different ways, and they say so in their own documentation. BrightLocal's Local AI Visibility report gives a score out of 100 that blends how often you are mentioned, where you appear, how positively you are described, and how strongly you are recommended. Local Falcon scopes its metric far more tightly:

Share of AI Voice (SAIV) is a metric that shows how visible your brand is within AI-generated search results for a specific query and area.

One is a blended index across platforms. The other is a per-query, per-area share. Both are useful inside their own tool, neither converts into the other, and a vendor telling you "your AI visibility score went up 12 points" has told you nothing you can audit.

I start with appearance rate because it is plain enough to check from the saved responses. Citation rate answers a separate question: how often the answer used the business or its site as a source.

A score without location detail misleads

A parent brand can post a healthy company score because a few mature branches appear often. That number can hide a new acquisition whose name, page, Business Profile, reviews, and citations still disagree with each other.

The operating view breaks the score down by location, buyer prompt, engine, competitor, and cited source. A plumbing branch can do well for drain cleaning and disappear for water-heater replacement. An HVAC branch can show up in Gemini and miss in ChatGPT. The company average does not tell the branch owner which problem to fix. The location-level audit guide shows how to store those dimensions.

What to do this week

Pick three branches and the five services that make each of them the most money. Write the question a customer would actually type for each service, without your business name in it, then run all fifteen on ChatGPT, Gemini, Perplexity, and Google AI Mode and record whether the branch was named. Divide the named answers by the total for each engine and put the four numbers next to the bands above.

If a branch is below 12.8% on an engine, check that its hours, services, and service area are right on the Business Profile and the directories before anyone writes content. If it sits near the 29.5% median, compare the answers it lost with the competitors and sources that appeared instead. If it clears 34.9%, protect what is working and look for the weak service line rather than chasing the company average. Keep the same prompts, engines, and markets every time you rerun, because changing any of them changes the denominator. If you would rather start from a measured baseline than a spreadsheet, run the grader on the weakest of the three branches first.

Methodology

The bands come from the corrected Cheers production panel pull of September 28, 2026: read-only queries over daily prompt aggregates dated August 6 through September 2, 2026. The panel is every company in the Cheers account tree with at least one organization categorized as home services, excluding Cheers-owned and internal research organizations and a Cheers test account. Each AI answer is counted once, using only the prompt-level row for each answer. Branded prompts, those whose text contains the company's own name or an approved alias, are left out. Each company's appearance rate is its named answers divided by all of its answers across every tracked location, non-branded prompt, and engine, and percentiles are taken across the 13 companies. Engines are ChatGPT, Gemini, Perplexity, and Google AI Mode. Published cells need at least ten companies behind them; no company is named. The rules are in the methodology.

Sources

The Cheers pages were checked on September 28, 2026, and the Semrush, SISTRIX, Local Falcon, and Google links on September 23, 2026. BrightLocal's help center blocked automated access on September 23; its definition was last checked on September 2, 2026.

• Home Services AI Visibility Index 2026. Cheers Research, edition September 2, 2026, corrected September 28, 2026. 13 home-services companies, 958,754 AI answers to non-branded prompts, aggregate only. Supports the percentile bands, the branded-prompt comparison, the engine medians and citation means, the ChatGPT position, and the prompt-shape rate.
• AI visibility statistics. Cheers Research. The same corrected panel figures, one dated statistic at a time.
• What is Visibility %?. Semrush Knowledge Base. The ranking-and-CTR definition quoted above.
• Calculation of the SISTRIX Visibility Index. SISTRIX, updated March 7, 2023. The search-volume and CTR weighting quoted above.
• What does each Local AI Visibility metric mean?. BrightLocal Help Center. Definitions of visibility score, mention rate, share of voice, and average position.
• How to use Share of AI Voice. Local Falcon Knowledge Base. The per-query, per-area definition quoted above.
• Optimizing your website for generative AI features on Google Search. Google Search Central, last updated July 10, 2026. A page must be indexed and eligible to show with a snippet before it can appear in Google's generative AI features.

Dylan Allen-Arnegård is the CEO and Co-Founder of Cheers. I help multi-location service brands get recommended by AI search and Google, with results tracked by location, employee, and competitor.