LLM vs Search Engine: Which Method Do Startups Favour?

When we wanted to find things out, a simple Google search would’ve been the way to go. Now, large language models such as OpenAI’s ChatGPT attract hundreds of millions of users each week. Demandsage data shows ChatGPT has 800 million weekly active users, double the number from February 2025.

Generative AI has become the way people find answers and it has even replaced standard research methods for many. Even with this growth, Datos research shows only 5.99% of desktop searches in July went to LLMs, which is more than double a year earlier but still far below Google or Bing.

 

Are Businesses Using LLMs And AI Tools?

 

Small businesses seem to be benefitting from AI Search features, though. Founder of Escape the Past, an independent escape room in EdinburghChris Wood says about Google’s AI Search:

“I was somewhat invisible on search engines until this new feature. Since July, we’ve seen a 36% increase in website traffic from organic search, and bookings are up significantly compared to the same period last year. It’s still early days, and at present Google doesn’t allow you to filter traffic statistics by AI Search, but the trend is encouraging. While it doesn’t replace our existing SEO efforts, we’re actively monitoring AI Mode results and optimising our content to make the most of it.”

“For small businesses like ours, it’s an early sign that quality and local expertise can now translate directly into visibility online. With all the concern around the impact of AI on smaller businesses, it’s refreshing to see AI search rewarding what we do best rather than what we spend on marketing.”
 

How Are Different Search Methods Performing On Mobile Vs Desktop?

 

AI search growth is happening mostly on desktop. BrightEdge data shows more than 90% of referrals from ChatGPT, Perplexity, and Bing are on desktop. Google is the exception, with 53% of its AI referrals coming from mobile.

AI Overviews appear 3 times more often on mobile ecommerce queries than on desktop. On desktop, they take up 80% more screen space and appear 39% more often. BrightEdge chief executive Jim Yu said businesses should design AI strategies with device context in mind, especially for shopping.
 

Do AI Search Tools Have Risks?

 
AI tools that are built into search engines can have faults as well, though. Last year, we had the scandal where Google’s new AI Overviews pulled in false information and presented it as fact.

A Columbia Journalism Review study found that many generative search tools, even premium versions, were sending users to inaccurate or syndicated sources rather than the original publishers.

This ends up misleading readers with an illusion of credibility and it undermines traffic to legitimate news outlets, threatening the financial stability of journalism.
 

What Are Startups Using More?

 

We’ve asked startups which method works for their companies. Here’s what they shared…

 

Our Experts:

 

Ben Duffy, Client Development Manager, Quirky Digital
Aaron Whittaker, VP of Demand Generation, Thrive Internet Marketing Agency
Aleksa Marjanovic, Founder & Marketing Director at Eternal Urns
Santiago Nestares, Co-Founder, DualEntry
Zeel Jadia, CEO, ReachifyAI
Monte Malukas, Founder, Peasy
Ignacio Merino, CEO, Freetour.com
Phillip Stemann, SEO Consultant, Phillip Stemann
Elie Berreby, Senior Organic Search Advisor, SEM King
Erling Løken Andersen, Hedge Fund Manager, Neverwinter Fund I SLP
Eric Turney, President and Sales and Marketing Director, The Monterey Company.

 

Ben Duffy, Client Development Manager, Quirky Digital

 

 

“It really depends on the industry and the audience. For younger, more tech-savvy markets, LLMs are becoming a powerful way to be discovered and generate traffic and revenue. On the other hand, if your target audience is older, search engines are still the go-to, as that demographic tends to place more trust in traditional search.”

 

Aaron Whittaker, VP of Demand Generation, Thrive Internet Marketing Agency

 

 

“Being a marketing agency, we’ve figured that search engines are still the most effective channel for bringing qualified traffic for startups since users go to search engines with PURCHASE INTENTIONS. When consumers come to Google, they are typically further down the funnel, searching for comparisons, pricing, or reviews.

“LLMs like ChatGPT, Gemini or Grok are great for EARLY RESEARCH and IDEATION, but they’re made more for discovery and education. As for cash-strapped startups, limited resources mean they should focus first on building visibility where the intent is already high, and avoid diluting them in scattered efforts.

“We had one B2B SaaS client, for instance, whose efforts on AI-driven content were huge but not resulting in conversions. When we transitioned to an SEO + paid search approach, leads increased over 40% in three months as we captured users at the purchasing decision point.

“We still recommend that clients experiment with LLMs for top-of-funnel content (like educating prospects), but Search is absolutely the answer when it comes to direct revenue growth.”

 

Aleksa Marjanovic, Founder & Marketing Director at Eternal Urns

 

 

“We rely on language models more often these days because they save us time synthesising and structuring information. When you’re running an ecommerce business, you’re constantly working across product research, customer communication and operational planning, which all lead to disparate data.

“Using an LLM, I can ask a question that bypasses extensive information and receive structured answers instead of sifting through various materials. What that clarity does is speed decision-making, especially when considering product opportunities or testing marketing approaches. At the same time, we don’t give up on search engines; they still serve an essential function as a verification against facts and as a way to extract raw, first-hand information that models cannot promise to prove for us.

“When I need flexibility, options, or to think strategically, I call an LLM. I was in the middle of a product expansion and had a long table of supplier data and regional regulations. I used the model to translate that into a useful comparison chart that helped my team balance cost against scalability.

“Search engines, by comparison, are my lifeline for specifics (such as shipping rates), checks on trademark use, and reviews of suppliers where I need direct links and primary sources. The division of responsibilities allows us to benefit from both perspectives. Discover provides the necessary information, while LLM transforms it into actionable tasks. That layering process maintains speed in our workflow, but more importantly, accuracy as we grow operations and reduce unnecessary errors.”

 

Santiago Nestares, Co-Founder, DualEntry

 

 

“Search is still the workhorse. Buyers Google their pain and land on your landing page if you’ve nailed intent. That’s compounding demand.

“LLMs are top-of-funnel distribution, not discovery. You don’t “rank” in ChatGPT the way you do in Google. Right now, they’re better for enablement — drafting, content ops — not for pulling in net new demand.

“So for startups: lean on search to get discovered, use LLMs to scale output. Long-term, LLMs may eat into search traffic, but today, if you’re B2B and you want customers, you still fight for search.”

 

Zeel Jadia, CEO, ReachifyAI

 

 

“AI is perfect for big picture research. Early on, it helps you cut through the noise and zero in on what actually matters.

“When the stakes are higher—metrics, key details, hard data—you need search. That is where you find verifiable sources to back up decisions.

“Both tools have flaws. Search can be gamed by SEO. AI is not immune to bias or manipulation either. The difference is, it is much harder to spot the noise in an AI answer.”

 

 

Monte Malukas, Founder, Peasy

 

 

“Being a small team, managing time effectively is critical for us. One feature we rely on heavily is deep research, whether through Chatgpt or Gemini. When a questionable topic comes up during team meetings (something that would usually require extensive googling), we simply ask the chatbot to perform deep research. Within a few minutes, we have a solid answer that often eliminates the need for follow up chats or additional calls to clarify the same topic. In other words we can discuss it real time. We apply the same approach outside of meetings, which saves us a lot of time on research tasks.

“Given the current limitations of Google search (although AI Mode is now also an option), we find this method not only faster but also more informative. AI chatbots provide clearer, more focused insights compared to the inaccurate hallucination sometimes produced by Google’s AI overviews. For up to date information like news, we still rely on search engines, though tools like Perplexity are starting to cover this gap by delivering fresher results and reducing the need to jump between platforms.

“We also use search engines when we know the answers require a wide range of sources or detailed analysis that AI chatbots may not fully provide. This happens especially when the questions are very specific or complex and we want to explore multiple perspectives beyond the limited options AI tends to show.

“Overall, we’ve noticed a significant decrease in our use of traditional search engines like Google because AI chatbots handle a large share of our research tasks with greater speed and relevance.”

 

Ignacio Merino, CEO, Freetour.com

 

 

“We don’t rely on just one method. At the moment, we mostly use search engines and PPC to attract customers. AI is still smaller for us, but it is useful for keeping people engaged and personalising their experience. I expect AI could work more like PPC in the future, so businesses should be ready. The key is to stay flexible and use each tool for what it is best at.”

 

Phillip Stemann, SEO Consultant, Phillip Stemann

 

 

“I spend maybe 95% of my time on search engines and 5% of my time on LLMs. LLMs still cover less than 1% of the traffic to my SaaS, and it’s the same I see for others in the industry; however, the leads that come from LLMs tend to have a higher conversion rate.

“A lot of the things I do for search engines also work for LLMs, so that’s why it’s a win-win, focusing on search engines.

“It’s all about getting mentions on the web and forums, building a strong backlink profile, and having content that answers the questions your potential customers might have for your industry.

“When I focus on LLMs, it’s long-tail questions I try to answer concisely on my website, whereas for search engines, it’s a bit broader, where I try to hit more broad topics.”

 

Elie Berreby, Senior Organic Search Advisor, SEM King

 

 

“Startups are by definition new “entities” that won’t appear out of the blue in AI answers.

“To become visible in LLMs as a new entity, a startup must first be discoverable by search engines such as Google and Bing. That’s because at this stage, the relationship between LLMs and search engines is circular.

“LLMs such as ChatGPT use web crawlers that scrape search engine results pages (SERPs) to find information on new or relatively unknown entities not already included in their existing training data.

“LLMs have been trained using a tremendous amount of information but when it comes to new entities, such as startups, this data is often outdated.

“And when content appears outdated, popular LLMs send their crawlers to search engine result pages.

“That’s why search engine indexing is the first step that allows crawlers used by AI models to detect new entities in the search results of traditional search engines.

“Once a startup’s content appears in search results and LLMs begin featuring the new entity in their answers, a high-quality digital PR strategy can reinforce organic visibility in AI-generated content.”

 

Erling Løken Andersen, Hedge Fund Manager, Neverwinter Fund I SLP

 

 

“At Neverwinter, we increasingly focus our marketing efforts towards large language models (LLMs) over traditional search engines. Google and similar search engines still have their place in the marketing mix, but their role has diminished. LLMs like ChatGPT are adaptive, conversational, and able to advise potential investors for our fund in a personal, direct way.

“For a hedge fund like Neverwinter, which operates an AI-driven crypto quant fund, the ability to interact semi-directly with users, albeit through ChatGPT, is far more aligned with how we approach new markets. In a way, LLMs can be compared to old school word of mouth: it’s personal, direct and trusted.

“This doesn’t mean that Google is “dead” – just that marketing has evolved beyond it. Google remains useful for discovery and certain fact checks. But for startups and funds that need agility, creativity, and real-time answers, LLMs are getting increasingly important.”

 

Eric Turney, President and Sales and Marketing Director, The Monterey Company.

 

 

“At The Monterey Company, we rely on both LLMs and search engines, but for different purposes. Search engines are still best when we need fresh, verifiable information, especially for fast-changing topics like supply chains, tariffs, or shipping updates.

“LLMs, on the other hand, excel at synthesising and contextualising data, whether that’s drafting content, building landing pages, or running scenario-based calculations that would take hours manually. For startups, I’d recommend treating them as complementary tools: search for facts, use LLMs for analysis and creativity, and let each play to its strength.”