Frequently asked questions

AI share of voice is the percentage of AI-generated answers, across engines like ChatGPT, Perplexity and Google AI Overviews, in which your brand is named or your domain is cited, relative to your competitors on the same set of prompts. Instead of tracking clicks or rankings, it tracks whether you show up inside the synthesized answer a user actually reads. A higher share means your brand is more often the one AI recommends when someone asks a buying-intent question in your category.

Market share measures revenue or unit sales across an industry; AI share of voice measures how often AI answer engines mention or cite your brand relative to competitors on the same prompts. The two are correlated but not identical — a smaller player with strong, well-structured content can out-punch a market leader inside AI answers if its pages are more citable and better aligned to the exact questions buyers ask. SoV is a leading indicator, not a lagging financial one.

Search behavior is shifting toward single synthesized answers instead of ten blue links, and a growing share of buyers act on that answer without ever clicking through to a website. If your brand is invisible inside those answers, you lose consideration at the exact moment of intent, and no amount of classic SEO reporting will show you the gap, because the click that would have logged it never happens in the first place.

A mention is when an AI engine names your brand in the body of its answer, which matters most for consumer brands, local businesses and hotels. A citation is when the AI engine lists your domain as a source or reference link, which matters most for content sites, SaaS products and anything competing on expertise. The AI Share of Voice Index tracks both signals separately so you can see which lever actually needs work.

The index tracks answers from ChatGPT, Perplexity, Google AI Overviews and Google AI Mode, with Claude and Microsoft Copilot coverage expanding as those platforms open up query access. Each engine has a different retrieval and ranking approach, so a brand can have a strong share of voice on one engine and near-zero visibility on another — which is exactly the kind of gap a single-engine check would miss entirely.

No. Share of voice measures how often you're mentioned or cited, not sentiment or how favorably you're described. A brand can appear frequently in comparison answers while being framed as the budget option or the runner-up. That's why the underlying data behind any share-of-voice number should always be inspectable — you want to know not just that you showed up, but how you were positioned relative to the brand AI actually recommended first.

For each tracked prompt, the index runs the same question repeatedly against each AI engine, records which brands are named and which domains are cited in the answer, and aggregates the results into a rate with a 95% confidence interval using the Wilson score method. Your share of voice is your rate divided across the full competitive set tracked on that prompt, rolled up across all prompts in a category to produce an industry-level index.

AI answers are probabilistic — the same question asked twice can return a different set of named brands or cited sources, because the underlying model samples its response rather than returning a fixed lookup. Running a prompt only once would capture a single roll of the dice and call it a measurement. Running it many times and aggregating the results produces a rate that reflects the real underlying tendency of the engine, not one noisy sample.

A Wilson confidence interval is a statistically sound way to express uncertainty around a rate calculated from a limited number of samples, and it behaves better than a simple percentage when the sample count is small or the rate is near 0% or 100%. Applied to share of voice, it means a reported number always comes with an honest range — so a 20% mention rate from 10 runs is flagged as far less certain than the same 20% from 200 runs.

Core index benchmarks are refreshed on a rolling monthly cadence, since AI answers drift as models update and as competitors publish new content, and a stale number would understate how quickly the landscape moves. Category and prompt-level detail pages show the date of their most recent measurement so you can judge freshness yourself rather than assuming every figure on the site was captured at the same moment.

Prompts are curated to reflect real buying-intent questions people actually type into AI assistants — comparison questions, 'best X for Y' questions, and direct recommendation requests — rather than being randomly generated or padded with irrelevant long-tail queries. Each prompt set is reviewed for relevance to the category it represents, because a share-of-voice number built on unrealistic prompts would tell you nothing about how your brand performs in the moments that actually matter.

Yes, and this is expected. A brand might be cited constantly on a technical comparison prompt because its documentation is thorough and crawlable, while barely appearing on a 'best for beginners' prompt because its content skews advanced. Share of voice is calculated per prompt and then aggregated, so the index intentionally preserves this variation instead of smoothing it into one flat number that would hide where the real gaps are.

The mechanics are the same everywhere, but what counts as a 'win' differs by industry. Consumer and hospitality categories care most about brand mentions, since a traveler asking for a hotel recommendation wants a name, not a source link. Content, software and services categories care more about domain citations, since being the linked reference on a technical answer drives both authority and traffic. The index reports both signals so each industry can weight the one that matters to it.

Categories with fragmented, review-driven purchase decisions — local services, hospitality, healthcare providers and B2B software comparisons — tend to show the widest gap between market-known brands and AI-cited brands, because AI answer engines favor structured, frequently updated, easily parsed content over reputation alone. A well-known brand with a thin or poorly structured website can score far lower in AI answers than a smaller competitor whose content is written specifically to be quotable.

E-commerce brands should weight product-comparison and 'best X' prompts heavily, since that's where AI answer engines most directly influence a purchase decision before a shopper ever visits a retailer's site. Because product pages change frequently, e-commerce share of voice is also more sensitive to staleness — pricing, availability and specification pages need to stay current, or an AI engine may quietly stop citing them in favor of a competitor with fresher data.

Very much so — arguably more than for content publishers, because local intent queries like 'best plumber near me' or 'top-rated dentist in [city]' increasingly get answered directly by AI assistants using a synthesized recommendation instead of a map pack. For these businesses, brand mention rate matters far more than domain citation, since the buyer usually just wants a name and a reason to call, not a link to click through.

B2B SaaS buyers frequently use AI assistants for comparison research — 'X vs Y' and 'alternatives to X' prompts — so citation quality and being named across many comparison variants both matter heavily. Because B2B sales cycles are longer, a single missed mention is less immediately costly than in consumer retail, but a sustained citation gap across a whole category of comparison prompts steadily erodes consideration before a prospect ever reaches a sales conversation.

Yes, and it happens often. AI answer engines reward clear, well-structured, frequently cited content more than brand size or ad spend, so a focused niche player with genuinely helpful, quotable pages can out-cite a much larger competitor whose content is thin or hard to parse. This is one of the more actionable findings the index surfaces: AI visibility is a content and structure problem you can fix, not just a budget problem you can't.

Start by identifying the specific prompts where you're absent but a competitor is consistently named or cited, then check whether your content directly and clearly answers that exact question. AI engines favor content that states a claim plainly, backs it with specifics, and is structured so the answer can be lifted cleanly. Rewriting a handful of high-value pages to directly answer your worst-performing prompts is usually faster than a broad content overhaul.

Yes — being crawlable is a prerequisite, not a nice-to-have. If AI crawlers like GPTBot, ClaudeBot and PerplexityBot are blocked by robots.txt, or your key content is hidden behind heavy client-side rendering, an AI engine simply cannot read it, let alone cite it. Fixing crawler access, clean HTML structure and fast page loads won't guarantee a mention, but it removes the most common reason a genuinely good page never gets a chance to be cited.

Yes. AI answer engines generally prefer sources that look current, especially for prompts involving pricing, rankings, comparisons or 'best of' lists, because a stale answer risks being wrong. Pages with visible last-updated dates, current statistics and regularly refreshed comparisons tend to get cited more consistently over time than pages that were published once and never revisited, even if the original content was thorough.

No — content that reads well for AI engines and content that reads well for humans overlap almost completely. Clear claims, direct answers near the top of the page, specific numbers instead of vague language, and logical structure with headers all help both a human skimmer and an AI retrieval system. Chasing AI-only tricks without serving real readers tends to produce thin content that neither ranks nor gets cited reliably.

Most brands see the first measurable movement within four to eight weeks of publishing or substantially revising content, since AI engines need to recrawl and re-index the updated pages before they can surface in a new answer. Improvement is also gradual rather than binary — you'll typically see mention or citation rates climb prompt by prompt rather than jumping all at once across your whole tracked set.

They help, but they aren't the whole story. Domain authority and a strong backlink profile make AI engines more likely to trust and surface a source, similar to how they influence classic search rankings. But AI engines also weigh how directly and clearly a page answers the specific question asked, so a lower-authority page that nails the exact prompt can still out-cite a higher-authority page that only addresses the topic indirectly.

CiteLens is the AI visibility measurement platform behind the data published on the AI Share of Voice Index. Where this index publishes free, public benchmarks by category and industry, CiteLens lets an individual brand track its own prompts, competitors and engines continuously, with a dashboard, confidence intervals and prioritized recommendations for closing specific gaps. Think of the index as the public map and CiteLens as the tool you'd use to actually navigate it for your brand.

Yes. You add your brand, your competitors and the prompts your customers actually ask, and CiteLens runs those prompts against AI engines on a recurring schedule, reporting your mention and citation rates with confidence intervals over time. Unlike the public index, which reports category-wide benchmarks, CiteLens gives you prompt-by-prompt detail specific to your business, including the actual AI answers behind every number.

Both. Alongside the raw measurement, CiteLens generates prioritized recommendations that flag which of your pages aren't structured to be citable, which specific prompts you're losing on, and which third-party sources you should be aiming to get mentioned on instead. The goal is to turn a number into a concrete next action rather than leaving you with a dashboard and no path forward.

Yes. A free GEO audit tool checks whether your site is even reachable by AI crawlers such as GPTBot, ClaudeBot and PerplexityBot, and flags structural issues that would prevent your content from being cited regardless of how good it is. Since being crawlable is a prerequisite for being cited, this free check is usually the right first step before investing in a fuller share-of-voice tracking setup.

Yes. CiteLens is built around comparative measurement — you add the competitors you actually care about, and every tracked prompt reports your mention and citation rate directly alongside theirs. This turns a vague worry like 'AI probably recommends someone else' into a specific, quantified answer: which prompts you're losing, to whom, and by how much, updated on a recurring basis rather than as a one-time snapshot.

No. CiteLens measures your AI visibility entirely from the outside by querying AI engines on your behalf — there's no code to install, no plugin, and no DNS change required to start tracking. Measurement begins as soon as you add your brand and a starting set of prompts, and any improvements you later make to your own site are reflected in your tracked scores automatically.

The data comes from live queries run against real AI engines, not synthetic estimates or third-party surveys. Each tracked prompt is submitted to the actual answer engine, and the response — including which brands are named and which domains are cited — is parsed directly from that live output. What the index reports is the answer a real user would have seen if they'd asked that exact question at that exact time.

On the public index, category pages summarize aggregated rates rather than showing every individual raw answer, to keep the benchmarks readable. For brands that want the underlying evidence — the actual answer text, named brands and cited URLs behind every number — CiteLens's citations explorer exposes that detail per prompt, so you can inspect exactly what the AI said rather than trusting a summarized figure alone.

The index tracks a curated set of buying-intent prompts across dozens of industry categories, each run repeatedly across multiple AI engines to build a statistically meaningful sample. The dataset grows over time as new categories and prompts are added, and each category page states how many prompts and runs its published figures are based on, so the underlying sample size is never hidden behind a single headline percentage.

Yes. The how-to-measure page on this site documents the full methodology in plain language — how prompts are selected, how many times each is run, how mentions and citations are distinguished, and how the Wilson confidence interval is applied. The goal is that anyone can audit how a published number was produced rather than having to take a black-box score on faith.

Share of voice is always calculated relative to a defined competitive set on a defined list of prompts, and that set is disclosed per category rather than left implicit. A brand isn't penalized for not being manually added to every possible comparison — it's simply not represented in categories where it isn't tracked. This is why the index states its competitive set and prompt list openly instead of implying a single universal ranking.

Yes. The index is actively expanding into new industry categories over time, and category requests from brands or researchers help prioritize which comparison sets get built next. Since useful share-of-voice benchmarks require a realistic, curated prompt list for a category to be meaningful, requests that come with example real-world buyer questions for that category are the most useful and fastest to act on.