I’ve often wondered how people are truly interacting with AI technology and what those interactions mean for our digital strategies. As I dive into recent survey data, it’s clear that real-world users are blending short queries with personal context, altering how brands achieve visibility in AI-driven searches.
Initially, I was surprised to learn that most people don’t use AI in the manner many Generative Engine Optimization (GEO) discussions suggest. Through surveys conducted by Stella Rising, where I’m the VP of SEO, we discovered that many AI prompts closely resemble traditional search engine queries.
For instance, in a beauty-focused study from August 2025 and a general study from January 2026, most prompts were succinct and keyword-driven, much like a Google search. However, many users are now providing AI systems with personal details, such as location and preferences, creating a deeper level of personalization.
Based on these findings, it’s evident that GEO strategies need to embrace this dual approach: accommodating classic keyword searches while optimizing for prompts enriched with personal context. This challenge presents a significant opportunity for brands willing to navigate this new landscape.
A lot of people are still typing like it’s 2008
A significant revelation from the surveys is that typical AI users still submit minimal inputs, hoping for optimal results.
Notably, our January general-audience study indicated:
- Two-thirds of users wrote prompts with 15 words or less.
- Only a small faction, about 12%, crafted what might be considered a comprehensive AI prompt.
- Most framed their questions while very few issued direct commands.
When I replicated a basic scenario — asking for a shoe recommendation — the average response consisted of eight words. Real entries included queries like:
- “Shoes nearby”
- “Tennis shoes”
- “Nike”
- “Ladies tennis shoes size 7 near me”
- “Best price for hiking shoes”
These align closely with findings from Semrush’s clickstream data, showing that the average prompt ranges between 4.2 and 8.7 words, paralleling standard Google queries. Structured, detailed prompts often surface in tasks beyond simple searches, like coding or content creation.
The shift between the two surveys
In the beauty-focused August 2025 survey, nearly half the prompts were firm, SEO-keyword-shaped. However, by January 2026, such prompts reduced to about 30%, with richer context becoming more prevalent.
Key observations included:
- Nearly a quarter incorporated the term “best,” highlighting an opportunity in “best [category]” visibility.
- A noticeable percentage mentioned budget or price, pointing to financially mindful consumers.
- “Near me” remained a common phrase, adapted from Google to AI interactions.
- A notable share included personal attributes, reinforcing the importance of personal context in queries.
However, the varying audiences surveyed offer caution. The 2025 beauty panel represented a unique demographic, while the 2026 group was more general and transactional, showcasing more complex query evolution.
The user embedding layer is where this gets interesting
The data revealing that 32% of users incorporate personal context into their prompts is significant. This includes details like job roles or life scenarios that traditional search queries do not capture. Real-world queries from users might include:
- “What shoes are ideal for standing all day at work?”
- “Find affordable running shoes on Amazon; size men’s 10.”
- “Suggest trendy, comfy women’s shoes, size 8 wide, under $120.”
The last example incorporates several layers of identity and specifics, which typical search engines never explicitly addressed. The embedding layer fuels AI’s ability to ‘know’ its user, leveraging past interactions to tailor responses, and it’s a game-changer for brand visibility.
Brands need to recognize that purchase-driving prompts often diverge from those seen in search engine results pages (SERPs). Real prompts hold significant buying influence and highlight the importance of context-rich brand mentions within AI interactions.
Where synthetic prompts fit — and where they don’t
Constructing synthetic personas helps test AI models’ representation of different user traits. However, synthetic prompts frequently miss the nuanced, ongoing dialogue a real user shares with AI tools. These personas can illuminate potential brand-user interactions but shouldn’t be the sole basis for measuring success in AI visibility.
Instead, complement synthetic prompts with insights from real user interactions for a holistic view. Pull real-world data from customer inquiries, support tickets, and search patterns to gauge true user engagement with your brand.
What to actually track
The current dynamics in AI search query patterns prompt us to reconsider our tracking strategies. With retrieval rates soaring, traditional SEO keywords are far from obsolete in AI contexts.
Yet, it’s crucial to focus tracking efforts wisely. Generic terms or single-brand queries may not yield insightful visibility information. Here’s how I recommend setting up an effective tracking framework:
- Use synthetic-persona prompts to cater to user embedding layers.
- Gather a set of real prompts from various data inputs for short, retrieval-invoking prompts.
- Maintain a qualitative set of context-heavy prompts to ensure content relevance and thoroughness.
What the broader data tells us about AI search
Further insights from January 2026 underscore why these prompt configurations matter in AI search:
Users increasingly trust AI recommendations
Approximately 68% of respondents trust AI recommendations more than Google’s, highlighting a trust transition driven by personalization and a lack of advertising clutter.
AI search is becoming a daily habit
Half of active AI users engage with these tools daily, gradually shifting dependency from Google to AI for common tasks. This shift signifies a change in how search habits are being shaped by AI convenience.
Citations still drive traffic
A substantial number of users still click on citations, validating that mentions within an AI context act as a gateway rather than an endpoint, showing the importance of monitoring and optimizing referral traffic through AI channels.
Voice may finally be having its moment
Voice interactions are finally seeing substantial usage, suggesting the long-predicted rise in voice-activated search is materializing, reinforced by the data from Ahrefs indicating visible shifts in clickthrough patterns.
In summary, AI search is taking form as a more personalized, interactive endeavor. It blends traditional intent with modern layers of user context, posing new demands and opportunities for content optimization. SEO and GEO strategies need to align closely with these evolving practices to maintain competitive edge.
What changes — and what doesn’t
As an SEO strategist, here are my top three recommendations for leveraging these insights:
- Revamp Your Prompt-Tracking Strategy: Blend synthetic prompts with real user inputs for a fuller understanding of AI visibility.
- Align Content with User Embeddings: Identify key user personas and ensure your content addresses their specific needs.
- Continue SEO-Keyword Optimizations: Traditional searches still play a crucial role, especially with high retrieval rates in play.
It’s vital to recognize that while AI evolves, many users still engage reminiscent of Google’s era, albeit within a platform more attuned to their specific contexts. This understanding guides where our optimization efforts must focus, staying attuned to changing user interactions and preferences.
Methodology
The studies referenced were spearheaded by Stella Rising. You can delve into them further in the report titled, “New Data: How Consumers Use LLMs for Search in 2026 (And What It Means for GEO).”
The August 2025 study surveyed 178 members of Stella’s community specializing in beauty, while the January 2026 survey covered a broader user base of 524 active users with some margin of error. These insights offer a directional lens into the broader adoption and interaction patterns within the AI space.
Inspired by this post on Search Engine Land.
Leave a Reply