It's common for brands to see referral traffic from ChatGPT, Perplexity, or other AI tools and assume this should translate into a high Gumshoe visibility score. However, traffic and visibility measure two completely different things.
This article explains why referral traffic does not necessarily indicate strong AI search visibility and how to properly evaluate what Gumshoe is showing you.
Traffic ≠ recommendations
Referral traffic from LLMs does not necessarily indicate that a brand is being broadly recommended across personas, topics, and models.
Traditional SEO tools measure “clicks,” not:
- How often an LLM recommends a brand
- Which personas the brand appeals to
- Which topics the brand appears for
- Whether the queries were commercial or informational
- Whether the behavior is consistent across all major AI models
If ChatGPT or Perplexity occasionally surfaces a link, often due to long-tail or informational queries, those clicks show up as referral traffic. That's positive, but it doesn't indicate meaningful visibility into recommendations for target personas.
Gumshoe's visibility score reflects structured, mid-to-low-funnel diagnostic tests, not incidental link appearances.
What generates traffic from informational queries?
LLM referral traffic is often generated by queries such as:
- definitions
- troubleshooting questions
- generic product lookups
- AI hallucinations or incidental citations
Gumshoe focuses on recommendation-style, commercially relevant prompts, the kinds of interactions that happen between AI models and your ideal customers. These are the queries that matter most for visibility and revenue.
A brand can receive LLM traffic and still have very low visibility if:
- The traffic comes from informational queries
- The traffic does not align with the personas or use cases that matter to the business
- The traffic appears only from one or two models
How do I validate a Visibility Audit when traffic data appears contradictory?
If you want to cross-check visibility, here are practical steps:
1. Check landing pages receiving LLM referral traffic
Then inspect them in Google Search Console or Bing Webmaster Tools.
Ask yourself:
- What keywords does this page rank for?
- Are those queries informational or commercial?
- Do they match the personas and topics used in the project?
Often, you'll find traffic is coming from queries that are not aligned with your business objectives.
2. Compare topics and personas
If the personas and topics in the project don't reflect the real business goals, your project needs tuning.
If they do align, the visibility score is accurate, regardless of incidental LLM traffic.
3. Start a second project (optional)
If you want to test whether visibility differs for queries that align more closely with your traffic:
- Harvest keywords or queries directly from GSC/Bing.
- Create a new project with a product category that encompasses those specific keywords.
- On the Topics & prompts page, edit the topics to reflect only the harvested keywords. Delete all system-created topics.
- Regenerate all prompts using the Actions menu in the top right.
Doing this and running the Visibility Audit might show higher visibility, but only if the newly generated prompts align with the audience your brand wants to attract.
If they don't, you've identified a mismatch between the traffic the brand is currently getting and the traffic the business actually wants.
When does traffic indicate a project needs tuning?
AI models do not always understand every industry correctly. Sometimes they surface topics or personas that don't reflect the business's goals.
To tune a project, check whether:
- Personas match the desired buyers or users
- Topics represent true commercial categories
- Prompts align with the real decision-making journey and are recommendation-oriented
If these elements are aligned, Gumshoe will accurately reflect AI visibility. If not, adjusting them ensures the model output becomes relevant and actionable.