Gumshoe Research is a curated section of the platform where we publish original data studies and ongoing analyses about how AI models cite sources across the web. It is separate from your reports and reflects patterns we observe across the broader Gumshoe data set, giving you industry-level context for what you see in your own results.
You can access it from the Resources section in the left navigation.
A Quick Note on Research Resource Update Frequency
Gumshoe's Research pages are published and updated less frequently than the core product. The data you see here reflects analyses run at a specific point in time and may not incorporate the most recent model versions, citation patterns, or platform changes. We update research studies when we have enough new data to make the findings meaningfully different from the prior version.
If something in a research page appears to conflict with what you are seeing in your own reports, your report data is more current. The research section is intended to provide directional context and industry-level patterns, not a live feed. We note the data period at the top of each study so you can assess how recent the underlying analysis is.
If you have questions about how to apply research findings to your specific account, feel free to reply to any support email or reach out at support@gumshoe.ai.
About the Citations Research Page
When AI models answer questions, they pull from external sources to support or ground their responses. Those sources are citations. Knowing which domains get cited, how often, and in what contexts gives you a window into how AI systems are forming answers in your category. This analysis draws on millions of citations collected across Gumshoe's data set, so you can see where influence is concentrated and where there may be room to build it.
Category Distribution
This bar chart shows how citations are distributed across content categories. Each bar represents a category (e.g., e-commerce, news, health, or technology), and its height reflects how frequently AI models cite domains in that category.
Use this chart to understand which types of content AI models favor when generating responses. This gives you insights into where your content publishing efforts are best focused.
Top Domains by Category
This section breaks down the most frequently cited domains within each content category. Each category has its own table showing the domains AI models most often reference, along with citation counts.
Reviewing this table tells you which source domains are dominant across all industries.
Top Domains by Model
These charts show citation patterns broken out by individual AI model. Each chart displays the domains most frequently cited by that specific model.
Different AI models draw from different sources and have different citation tendencies. A domain that ranks highly for Google may not appear at all for OpenAI.
SERP Position Histogram
This chart shows where cited domains tend to rank in Google search results, broken out by AI model. The x-axis groups citations into position ranges (1-3, 4-10, 11-20, and 20+), and the y-axis shows the percentage of citations that fall into each range. Each bar color represents a different AI system.
The chart shows whether a given AI model tends to cite sources that rank highly in traditional search results or draws from content further down the results page. A model with most of its citations clustered in the 20+ position range is pulling from sources that Google does not necessarily surface at the top, which means strong Google rankings alone do not guarantee AI citation.
You can use this chart to understand the relationship between traditional SEO performance and AI citations. If the models most relevant to your audience tend to cite lower-ranking pages, that is a signal that topical authority and content depth may matter more than raw search position when it comes to earning AI visibility.
Citations in Timeline
This line chart tracks how citations have accumulated over time across each AI model, expressed as a cumulative percentage. Each line represents a different AI system, and the steeper the curve at any point, the faster citations were being added during that period. Flat or gradual sections indicate slower growth.
The sharp upward curve tells an important story about content freshness. The vast majority of citations across all the AI models shown here come from relatively recent content, indicating that AI systems are not drawing equally from everything ever published. They skew heavily toward what is current. Content that was created or significantly updated in the last two years accounts for a disproportionate share of citations compared to older material that may be sitting unchanged.
For the GEO strategy, this is one of the more actionable signals in the research section. Keeping your key pages fresh, updating existing content with new information, and publishing consistently are not just good SEO habits; they are essential to your success. They directly affect whether AI models consider your content worth citing. An older page that hasn't been touched in years is at a significant disadvantage compared to newer content on the same topic.
How This Relates to Your Reports
The data in Gumshoe Research reflect patterns across the broader dataset, not your specific report. Think of it as the market context for your brand's performance.
About the Content Retrievability Page
Content Retrievability is a tool that tests whether a specific URL is part of an AI model's retrieval corpus, meaning whether the model will actually cite that page when answering relevant questions. It works by automatically generating targeted questions from your content, then querying the selected AI model to see whether your URL appears as a source in the responses.
To use it, enter a target URL and select an AI search model from the dropdown, then click Run Analysis. The Results Summary shows four key outputs: Status (whether the URL was successfully retrieved), Questions Generated (how many test questions were created), Citations Found (how many of those questions returned your URL as a source), and Citation Rate (the percentage of questions where your URL was cited).
Below the summary, you can see which specific questions cited your URL, browse the full list of questions that were tested, and review the Top Cited Sources table, which shows which URLs from your domain received the most citations and across how many questions.
Use this tool to diagnose why a specific page is or is not appearing in AI responses. A low citation rate on an important page is a signal that the content may not be structured, written, or indexed in a way that AI models can effectively retrieve.
Persona Cloud
Persona Cloud is an interactive 3D visualization of all personas in the Gumshoe system, plotted using UMAP dimensionality reduction. Each dot represents a persona. Blue dots are AI-generated personas and green dots are human-created ones. Conceptually similar people appear closer together in the visualization.
You can rotate and zoom the 3D chart using your mouse. Clicking on any dot reveals the persona's name and its UMAP coordinate values. The Quick Examples bar at the top lets you jump to a few sample personas to explore how the system works.
Click on the dots and review the persona details and neighbors below.
This view is useful for understanding the range and distribution of buyer personas Gumshoe uses across the platform. It can help you evaluate whether the personas in your own reports are well differentiated or cluster closely with others, which may suggest they are generating similar prompts and producing redundant data.
Persona Survey
Persona Survey is an experimental analysis tool prototype that shows how AI models respond to purchase-intent questions when filtered through the lens of a specific persona. It presents a baseline AI response before any persona context is applied, then shows how the response and its sentiment shift when the persona is introduced across each supported model.
Each model section displays the LLM's response after persona exposure, a sentiment distribution chart showing how the response skews across a five-point scale (from Definitely Not to Definitely Yes), and an Expected Value score out of 5.00. A positive change value next to the model name means the persona context made the AI more favorable toward the answer. A negative value means it made the response less favorable.
Review this feature and let Gumshoe know if having a similar tool within the Gumshoe product would be helpful. What else would you like to see? How would you change it?