A SaaS marketing lead runs the same brand through three tools in one afternoon: Goodie’s Brand Command, Profound’s citation-share dashboard, and a free AI visibility checker found through a quick search. The numbers do not agree. One says 34% citation share. Another says 12%. The free tool says 61%. Nobody is lying. Each tool is measuring a different slice of a different data source, and that gap is worth understanding before anyone reports a number to a VP.
Key Takeaways
- Citation percentages from different AI visibility tools are not measuring the same thing, even when the label says “citation share.”
- The gap comes from three variables: which models get queried, how many prompts are sampled, and what counts as a citation.
- Share of citation and share of voice are different metrics, and a tool that blends them will always report a higher number than one that does not.
Why the Same Brand Returns Three Different Numbers
Every AI visibility tool does the same three things in some order: it sends prompts to one or more models, it reads the answers, and it counts whether the brand shows up as a cited source. The percentage it reports depends entirely on how it does each of those three things.
Change the prompt set and the number changes. Change which models get queried, ChatGPT versus Gemini versus Claude, and the number changes again, because each model pulls from a different base. ChatGPT draws from across the open web. Gemini cross-references Search, YouTube, and Scholar. Claude leans on high-authority publications and documentation. A brand strong in documentation and weak in social content will score well on Claude and poorly on Grok. A tool that only queries one or two models will report a number that reflects that bias, not the brand’s actual standing.
What Goodie’s Brand Command Is Actually Counting
Brand Command runs a defined set of prompts against multiple models on a recurring schedule and logs whether the brand appears as a named, cited source in the response, not just mentioned in passing. That distinction matters. A model can describe a product category and happen to use a brand name without treating that brand as a source it is pulling information from. Brand Command is built to separate the two.
Because the prompt set is fixed and repeated, the dashboard can show trend lines week over week. That is useful for tracking whether a specific campaign or content push moved the needle. But it also means the percentage is only as representative as the prompt list behind it. A brand tracked on fifty branded and category prompts will post a different number than the same brand tracked on five hundred.
What Profound’s Citation-Share Dashboards Pull From
Profound builds its citation share around a defined competitive set within a category, then measures how often the brand gets cited relative to named competitors across that same prompt pool. The number it reports is relative, not absolute. A brand can hold a strong citation share inside its own dashboard and still get cited less often in raw terms than a brand that falls outside the competitive set Profound built for it.
This is where the compounding effect of citations starts to show up in the data. Once a model cites a brand for one query, the next related citation becomes more likely, because the model has effectively learned to associate that brand with the topic inside the session. A dashboard sampling the same prompt cluster repeatedly will capture that compounding and report a rising share. A one-off free checker running a single prompt will not.
total AI citations recorded for Moburst’s own brand, with 309 unique pages cited and an average cited position of 1.77.See the case study
A number like that only means something because it comes with context: which pages got cited, where they ranked, and how the volume moved month over month. A single citation percentage, stripped of that context, tells a reader almost nothing about whether the trend is improving or which content is doing the work.
Why Free AI Visibility Checkers Skew High or Low
Free tools tend to run a handful of prompts, sometimes just one, against a single model, then report whatever percentage falls out of that tiny sample. A brand that happens to get cited on the one prompt the tool ran will show a misleadingly high score. A brand whose best citations live on prompts the tool did not ask will show a misleadingly low one.
Many free checkers also count any mention of the brand name as a hit, regardless of whether the model treated it as a source. That is why free tools often report higher numbers than paid dashboards built on the stricter definition of a citation: a model saying “brands like X and Y” is not the same as a model citing X as the source behind a specific claim. A loose counting method treats them identically.
A citation percentage is only ever as honest as the prompt set and the model list behind it.
Share of Citation Is Not Share of Voice
Share of voice measures how often a brand gets mentioned across the web overall. Share of citation measures how often AI models pull that brand in and name it as a source. These are not interchangeable. A tool that blends them will always produce a flattering number.
A brand can run a large press campaign, generate hundreds of mentions across news sites and social posts, and still show near-zero share of citation if none of that coverage is structured or current enough for a model to treat as a reliable source. Press coverage drives share of voice. Structured data, consistent entity references, and content that models can verify against other sources drive share of citation. A checker that counts both the same way is measuring awareness, not retrieval.
What to Do With Three Conflicting Numbers
Do not average the three percentages and report the mean. Instead, ask each tool three questions before trusting its output: which models did it query, how many prompts, and does it count a loose mention the same way it counts a sourced citation. Write those answers down next to the number.
What’s the actual goal here? If it’s a weekly trend line to show whether content work is moving the needle, a fixed-prompt dashboard like Brand Command is built for that. If it’s understanding standing against named competitors inside a category, a relative tool like Profound answers a different question entirely. A free checker is fine for a quick gut check on a single prompt, but it should never be the number that goes into a board deck. Run an AI Visibility Audit alongside whichever dashboard is already in use, and treat the audit as the baseline the recurring tool gets measured against, not the other way around.
FAQs
Why Does My Citation Percentage Differ Between Tools?
Each tool queries a different set of models, runs a different number of prompts, and defines “citation” differently, some count any brand mention, others only count the brand being named as a sourced reference. Those three variables alone can produce very different percentages for the same brand.
Which Number Should I Report to Leadership?
Report the number alongside its method: which models, how many prompts, what counted as a citation. A percentage without that context is not useful for tracking progress or comparing against competitors.
Are Free AI Visibility Checkers Worth Using At All?
They are useful for a quick directional check, but their small prompt samples and loose counting methods make them unreliable for trend tracking or competitive comparison. Use them as a starting point, not a source of truth.
Is Share of Citation the Same As Share of Voice?
No. Share of voice counts mentions across the web generally. Share of citation counts how often AI models name the brand as a source in an answer. A brand can be high on one and near-zero on the other.
