The rise of advanced conversational tools, spearheaded by platforms like ChatGPT, has drastically reshaped many assumptions within the digital marketing industry. For years, the prevailing consensus among SEO professionals has suggested a fundamental shift away from traditional, keyword-based searches, especially concerning local service providers.
The hypothesis was straightforward: as users increasingly interact with Large Language Models (LLMs), they would naturally adopt conversational prompts—asking full, complex questions rather than typing short, choppy keyword phrases. This perceived evolution fueled predictions that traditional keyword research and tracking, long the bedrock of search engine optimization (SEO), would quickly become obsolete.
However, recent observational data challenges this widespread assumption, particularly in the realm of local, transactional intent. A study conducted by observing everyday users utilizing ChatGPT to find professional local services—including healthcare providers and aesthetics practices—revealed a surprising adherence to established search habits.
The core finding is unambiguous: the vast majority of users, even when starting their journey on a cutting-edge generative AI platform, still rely on familiar, keyword-driven queries to connect with local businesses. This discovery has profound implications for how marketers approach local SEO and the emerging discipline of Generative Engine Optimization (GEO).
Challenging Assumptions in the AI Era of Search
Before the widespread adoption of tools like ChatGPT, the primary search entry point was Google, where keyword optimization dominated. With the advent of generative AI, the industry began to postulate a future defined by dialogue. The theory held that if a user was given the capacity for a full conversation with an AI model, they would utilize that capacity, especially for complex or high-stakes local needs, such as finding a dentist or a reliable chiropractor.
The observational study sought to validate or disprove this transition by placing real users in a natural search environment. Participants were explicitly asked to initiate their search for local service providers on ChatGPT and proceed as they normally would, which included checking websites, analyzing social profiles, and reviewing customer feedback. The goal was to answer critical questions about modern user behavior:
- Are customers engaging with ChatGPT conversationally when seeking local services?
- Has the intent to find local services fundamentally abandoned keyword-style searches?
- Is extended, multi-turn conversation common when the user’s ultimate goal is transactional (i.e., booking an appointment)?
The resulting data offers compelling evidence that, despite the technological shift, human behavior remains remarkably consistent, particularly when the search intent is to complete a tangible transaction.
The Enduring Relevance of Keyword Searches: The 75% Metric
One of the most significant findings of the observation was the high rate of traditional keyword usage. Across all observed sessions where users searched for local services, a remarkable 75% included at least one prompt that would be classified as keyword-based.
This runs directly counter to the narrative suggesting that conversational prompting has fully superseded short-tail and geo-modified queries. For many digital marketers who have been tracking keywords for decades, this data provides a vital reassurance: the foundational principles of SEO are still active, even within the confines of a sophisticated LLM interface.
Old Habits Die Hard: Efficiency in Transactional Intent
The primary driver behind this continued reliance on keywords appears to be efficiency. When a user has high transactional intent—meaning they need a specific service provider, like a “dentist in Chicago” or “dentists montgomery”—they gravitate toward the shortest path to the desired result. Providing the full address and service type in a concise format often yields the necessary list of recommendations quickly.
Consider the effort required. It is demonstrably simpler and faster to input a concise query like, “dentist 11214” or “good plastic surgeons in brooklyn 11214 area” than to construct a long, descriptive sentence such as, “5 good dentist according to online recommendations near india street, brooklyn, new york.” This pattern of behavior highlights a fundamental principle of digital interaction: users will almost always choose the lower-effort option if it delivers the required information effectively.
In the context of local services, the user’s primary concern is obtaining contact information, location details, and reputable recommendations immediately. The conversational aspect of the AI is secondary to the utility of the list it generates.
Implications for Generative Engine Optimization (GEO)
This finding mandates a revisit of strategic discussions surrounding Generative Engine Optimization (GEO). Some proposed GEO models included a mandatory step where transactional keywords were fed into a separate tool to convert them into longer, more natural language sentences before being tested in the LLM.
The study suggests that for local services, this conversion step is often unnecessary and potentially inefficient. Since users are already entering keyword-centric prompts, optimization strategies should focus on ensuring that local business data (NAPs—Name, Address, Phone—and service descriptions) are robust and clearly associated with these core keywords and geo-specific modifiers.
The fact that users are still entering phrases similar to “dentist in chicago” means that local keyword research and tracking remain highly valuable in the generative AI era. SEO professionals must continue to monitor the performance of these core terms to understand user demand and competition, even if the result is delivered through a chat interface rather than a traditional Search Engine Results Page (SERP).
Local is Not that Conversational: The Low Prompt Count
Beyond the persistence of keywords, the study uncovered another critical fact about user interaction with ChatGPT for local needs: the sessions are rarely characterized by extensive, back-and-forth dialogue.
The data shows that nearly half of the sessions—45%—were concluded after a single, “one-shot” prompt. This means the initial query provided sufficient data for the user to transition to the next step, which typically involves visiting external websites, checking reviews, or calling the recommended businesses.
Furthermore, when follow-up prompts did occur, they were often simple iterations rather than deep conversational engagements. A full 34% of second prompts were merely requests for more results (e.g., “Give me five more options” or “Show me someone closer”).
Average Prompts per Local Task
When searching for local services, the average ChatGPT user employed only 2.1 prompts per session. This low number underscores the transactional and utilitarian nature of these interactions. Users are looking for quick answers and actionable results, not philosophical discourse.
The average number of prompts varied slightly depending on the complexity or perceived risk of the service being sought:
| Task | Average number of prompts |
| Find a new dentist | 2.41 |
| Find a place to get botox | 1.96 |
| Find a dermatologist to check a mole | 1.71 |
| Hair transplant | 1.33 |
| Find a chiropractor | 2.33 |
| Decided to get a facelift | 2.00 |
Interestingly, high-involvement services like finding a new dentist or a chiropractor showed slightly higher conversational depth (2.41 and 2.33 prompts, respectively), potentially reflecting a need for more refinement or specific criteria (like insurance compatibility or specialized services). Conversely, targeted medical searches, such as seeking a dermatologist for a mole check (1.71) or specific surgical procedures like hair transplants (1.33), tended to be faster and more direct. This suggests that when the service need is highly specific, users know exactly what they are looking for and require less initial guidance from the LLM.
While the potential for conversational interaction is technically present in ChatGPT, the practical use case for local, high-intent searches minimizes the dialogue. The model is treated less like an oracle and more like a highly efficient search refinement engine.
Delineating Intent: Transactional vs. Informational Use
The distinction between *transactional* intent and *informational* intent is key to understanding these findings. Anecdotal evidence strongly suggests that when people use ChatGPT for purely informational purposes—such as summarizing complex topics, brainstorming ideas, or asking broad “how-to” questions—they engage in much longer, more robust conversations.
When the intent is transactional, however, the goal is not learning; it is *doing*. The user needs to move quickly from the search interface to a business’s website, phone number, or physical location. In these scenarios, extensive conversation becomes a barrier to efficiency.
This difference in prompting behavior based on intent means that local SEO strategies must be decoupled from general content marketing strategies for generative AI. While blog content may need to be structured to answer deeply conversational queries, local business data must be optimized to respond instantly and accurately to concise keyword prompts.
The Future Trajectory of Local AI Search
While the current data confirms the dominance of keyword-based transactional searches, it is crucial to acknowledge that user behavior is dynamic and may evolve over time. Several factors could shift the prompting landscape:
Increased User Acclimation: Many participants in the study were using the free version of ChatGPT and may not have been fully accustomed to leveraging its conversational capabilities. As users become more fluent in prompting sophisticated LLMs, their reliance on keywords might decrease.
Integration and Plugins: If LLMs like ChatGPT become more integrated with mapping services, real-time availability, or booking systems, users might naturally start asking more complex, multi-variable questions (e.g., “Find me a highly-rated chiropractor open next Tuesday who accepts Cigna insurance”). Currently, the keyword prompt is often the best way to extract a simple list of candidates before the user conducts the final vetting.
Ease of Search: Ultimately, the model’s capacity to deliver accurate results from short prompts is its own greatest constraint on conversational growth. If typing “dentist 11214” reliably provides the necessary information, users have no incentive to type more.
For the foreseeable future, however, the SEO strategy for local service providers must be built on the foundation of current user behavior: efficient, keyword-driven transactional searches.
Strategic Takeaways for Local SEO Professionals
These findings provide crucial guidance for local SEOs adapting to the AI-first search era.
Reaffirming Keyword Research and Tracking
The notion that traditional keyword tracking for local businesses is irrelevant has been strongly refuted. Professionals must continue to prioritize comprehensive keyword research, especially focusing on:
- Geo-Modifiers: Integrating city, neighborhood, zip code, and “near me” language into optimization efforts.
- High-Intent Phrases: Ensuring content and structured data correctly target terms like “best [service] near me,” “[service] cost,” and “[service] reviews.”
- Service Specificity: Understanding how users search for niche services within the larger category (e.g., “pediatric dentist” vs. “emergency dentist”).
If 75% of users rely on these keywords, tracking their performance across standard search engines and anticipating how LLMs interpret them is paramount.
Optimizing Local Infrastructure for LLM Extraction
Since the user engagement with ChatGPT is often short and transactional, the primary goal of the local business is to be present in the AI’s initial, concise response. This requires rigorous optimization of underlying local SEO signals:
Consistent NAP Data: Ensuring Name, Address, and Phone number are uniform across all platforms (Google Business Profile, directories, and website schema). Inaccurate or conflicting data will hinder an LLM’s ability to confidently recommend a business from a simple keyword prompt.
Structured Data Markup: Implementing robust Schema Markup (especially `LocalBusiness` and relevant service types) allows LLMs to easily ingest and verify key business details. LLMs rely on well-structured, authoritative data to formulate their definitive, short answers.
Review Signals: Because users are looking for established credibility when seeking local services, strong, recent reviews are essential. An LLM’s recommendation is often tied to perceived quality, which is frequently inferred from aggregated review data.
The findings from this observational study provide a necessary dose of realism to the excitement surrounding generative AI. While the technology promises infinite conversational depth, human users, when faced with a transactional task like finding a healthcare provider, prioritize speed and efficacy. For local SEO, the core objective remains the same: ensure the business is the most relevant, reputable, and easily verifiable answer to a focused, keyword-driven query.
The full study and associated data provide an in-depth look at the specific prompts and user flows observed during this research into how ChatGPT users navigate the search for local services. Ultimately, the data confirms that adapting to the AI future means merging traditional, highly-effective keyword strategies with modern, structured data optimization.