Showing Up In AI Search Isn't The Same As Getting Recommended

Direct Source Verification: This story is aggregated from Forbes (forbes.com). Full reporting rights and copyright belong to the primary publisher.
Andrew Higgins is CEO of Parsnipp, an AI search marketing company.

Andrew Higgins is CEO of Parsnipp, an AI search marketing company.

getty​For the last 12 months, a lot of marketers have been diving into AI search marketing, and most have been defining success with a new KPI: “AI visibility.”

AI visibility has been our primary measure of whether a brand is succeeding in AI search and appearing in LLM-generated answers related to its brand and market. But what we’ve learned at Parsnipp is that visibility tells only part of the story. There’s a much larger conversation (pun intended) happening around how people interact with AI, and it’s one we need to start paying closer attention to.

When we started building in AI search, we measured what everyone measures: how often a brand shows up in an AI answer. We tracked mentions, citations, share of voice, visibility scores, etc. A lot of this is derived from traditional search, and it’s the habit that’s trained into all of us. Win the keyword—or in this case prompt—and move on to the next.

The more we tested and replicated the ways people actually use AI, the more it changed how I thought about the channel and what it means for brands.

We wanted to see what happens after the first answer, because while most tools measure a set of prompts and AI responses for visibility, that’s not an accurate reflection of how real people use answer engines. So instead of firing single prompts at a model, my team simulated full shopper conversations: a person with a need, follow-up questions and constraints that come up along the way (i.e., the way people actually talk to LLMs).

The scenario was simple: headphones under $150. We ran 320 conversations across ChatGPT and Gemini 3.8 Flash in two independent batches of 160 to check if the pattern held. What we found was that across 960 brand mentions, only 700 became an actual recommendation. Nearly three in 10 mentions went nowhere. This means what marketers considered a “win” was actually far from it.

We had brands that scored incredibly well on visibility but almost never were the product or brand actually recommended by the end of the interaction with the shopper. On a visibility dashboard, that looks like a strong brand, but it was missing the last two-thirds of the shopper’s journey through AI.

In traditional search, appearing in results was the first step in the customer journey. From there, our website and retail partners took over the consideration phase, helping customers compare products, read reviews, shop for the best price and, ultimately, make a purchase.​

But in today’s world, with so much of that interaction taking place on one surface, between a shopper and their agent, AI now folds much more of this funnel into one channel. So, we’ve started thinking full funnel again, just inside the AI Search experience:

• Mentioned: The assistant brings you up.

• Considered: You stay on the shortlist as the shopper adds needs.

• Recommended: You’re the pick, not the runner-up.

• Chosen: The shopper, or their agent, buys.

Each stage leaks for a different reason, and most brands are only measuring the first one.

You don’t need special software for a first pass. Here’s the process I’d hand any marketing team:

1. Start With One Category Of Prompts: Pick the one that drives the most revenue, and don’t try to map everything at once. Maybe it’s your competitive prompts, broad industry prompts, etc. ​

2. Write Briefs, Not Keywords: List 20 to 30 real buying situations with the constraints included. “Running shoes” are keywords whereas “moderately frequent runner with wide feet that is over 35” is a brief.​

3. Script The Conversation: Simulate the conversation broadly, then add constraints one at a time (budget, use case, fit, deal breakers). I’d also switch around the order of questions, too. As an example, in our test, results shifted depending on whether the budget came first or last.​

4. Run It More Than Once, On More Than One Model: Answers vary run to run as LLMs still have a fair bit of drift response to response. Running multiple batches is what told us our pattern was real.​

5. Log Two Numbers Per Brand: Start off simple. You don’ t need to track too many points along the journey at first. Even jumping from 1 (visibility) to 2 (visibility + mention) will make a huge impact. This means you’ll be measuring how often your brand is mentioned and how often it’s recommended. You can do it for your competitors too in the same research. Recommendations divided by mentions is your recommendation rate.​

6. Find Your Sweet Spot: Being both mentioned and recommended is the ideal position. If a brand is rarely mentioned but consistently recommended when it does come up, that’s primarily a discovery problem—and generally the easier one to fix. Neither position is ideal, but the more concerning scenario is a brand that’s frequently mentioned but rarely recommended. It creates the appearance of visibility without actually translating into preference or consideration.​

7. Identify What’s Holding You Back: Whether it’s price, product fit, a missing specification or a use case your brand doesn’t adequately address, identify the factors causing AI to steer customers elsewhere. These gaps are likely where you’re losing sales—and where your AI search marketing efforts should focus.

8. Fix The Evidence, Then Re-Run: Make your product pages, reviews and third-party coverage answer that gap directly since LLMs have a tendency to repeat what they can find. Then test again next month because models change super fast.​

If your AI search report only counts mentions, it’s telling you half the story—and maybe the wrong half. Count the mentions, track the picks, fix the gap, and start thinking of AI search as a full funnel marketing channel instead of just a visibility goal.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?

Original Source
https://www.forbes.com/councils/forbestechcouncil/2026/11/19/showing-up-in-ai-search-isnt-the-same-as-getting-recommended/
Visit Forbes ↗
SHARE STORY:
𝕏 f in

Related Coverage in Technology