Executive Summary: This article argues that the LLM Perplexity has captured market share by adopting “thick” positioning as a purpose-built, retrieval-first, research tool, in contrast to the “thin” general assistant LLM positioning strategy that defines most AI products.
LLMS Are Generally Marketed As General Assistants
A key selling feature of large language model tools (LLMs for short) like ChatGPT, Claude, Gemini, Grok, Copilot and Perplexity, is their ability to summarize and categorize information. To simplify, perhaps too much, LLMs are essentially research engines for knowledge work: they take in text, recall or look up related information, and then synthesize it into useful outputs, with many of their other notable skills emerging from this same process.
LLMs are faster at parsing and generating many kinds of information than humans, and so there is an obvious incentive to make use of them where reasonable. If you need the gist of a lot of information, they can summarize it faster than you can manually. While there are other use cases for LLMs, this is easily the skill they are most known for seemingly surpassing humans in.
Yet, oddly, most LLMs are marketed as general assistants, not as “thick” specific tools with clearly defined limits. The promise they will change how we do everything is their supposed selling point. They are not marketed as hammers built to work with nails; they are “thinly” positioned as genies that can easily, cheaply, and reliably grant any wish you can dream of.
Reality betrays this LLM positioning strategy marketing.
One of the best known issues with many LLMs is that they “hallucinate” information, which, crudely speaking, means they seem to brazenly make stuff up. Any well-informed human is therefore skeptical about LLM’s outputs. If summarizing or categorizing lots of information with any appreciable degree of confidence matters (most obviously in areas like medicine, engineering, law, etc) using time-tested methods is the prudent approach, not unreliable LLMs.
In other words, general assistant-style LLMs can appear to summarize and categorize technical material like a diligent graduate student, but their research habits are closer to those of a precocious kid with unrestricted internet access.
LLM Positioning Strategy Marketing:
Gemini, Copilot & ChatGPT Float By on Brand Recognition
Gemini, Copilot, and ChatGPT have the thinnest positioning strategy marketing as general assistant LLMs.
Gemini is named after the Latin for “twins,” which partly refers to Google internally merging its Google Brain and its Deep Mind to create the LLM, and partly refers to NASA’s project Gemini. So, if you don’t know the history of Google’s business, or American spaceflight, the name carries no meaning for you. I find it odd they kept the internal project name as the public-facing brand, because it’s not a name that will evoke meaning to people outside Google. In practice, Gemini primarily positions itself by easily integrating with existing Google software.
Copilot is the same story but with Microsoft’s software suite. There’s no history behind this name. Presumably “Copilot” was chosen to bring to mind a co-pilot who helps you fly a plane. (While subtly suggesting that it is also qualified to fly planes and do your work on its own). To be fair, this is likely a more evocative name than Gemini for most, but use cases are still unclear. Microsoft, like Google, can get away with this generic branding for now, but only because they are industry leaders. Thin positioning is less fatal when you monopolize a market.
Thus, both Gemini and Copilot are marketed as general assistants with no specific use cases beyond integrating with software individuals and businesses already use. They easily appear more like add-ons to existing services rather than valuable products of their own. That is not an insignificant feature, as it enhances the value of their software suites. But it is not a specific use case either, and that naturally makes it harder to sell these LLMs to informed users. (This would seem to be a pressing issue for tools that are primarily known for informing users).
ChatGPT in contrast, is named after its own underlying technology. It is a Chat Generative Pretrained Transformer. Crudely speaking, that means it’s a tool pretrained on data that can generate conversations that seem conversational. Again, it’s branded oddly. Rather like Gemini, it has a fine name for an internal project, but a bizarre one for a public facing product.
By comparison, Chat GPT is even more thinly positioned than Gemini and Copilot. It can’t rely on a monopoly like Google or Microsoft, or native integration with their suites. It is just marketed as an all purpose chatbot. This leaves it with the vaguest product identity by far. This is likely why, despite its early to market advantage, it is losing market share to other LLMs. Until it embraces an LLM positioning strategy marketing niche, this will likely continue. Being most known for excellence at nothing in particular is hardly a winning proposition.
LLM Positioning Strategy Marketing:
Claude & Grok Have Personality & Politics
In contrast, Claude and Grok are positioned more thickly, by politics and personality.
Claude is named after Claude Shannon, the “father of the information theory.” You will never have heard of him, unless you’re into the history of mathematics and computer science. This name choice is poor for LLM positioning strategy marketing with mass-market audiences.
To make up for that deficiency, Anthropic’s “Keep Thinking” and “Claude is a space to think” campaigns frame Claude as a collaborator for problem solving. Simultaneously, Claude is sold to users on the basis of its political stance: its behavior is constrained by a “Constitution” inspired by various liberal democratic and human rights norms and values. So, in practice, Claude sells itself to a particular politics and personality, but not a specific use case.
Grok is named after a concept popularized by Robert Heinlein in Stranger in a Strange Land to refer to a visceral and total understanding of something. This is evocative branding. Grok has also been characterized by its owner, Elon Musk, as “maximally truth seeking” and not “woke” compared to other LLMs.
This positioning is just as vague as Claude’s, though. Whatever “the truth” turns out to be, it has no logically necessary relationship to whether it is “woke” or not. Anyone committed to seeking truth knows it cannot be defined by rejecting an ideology in advance. So, in practice, Grok’s brand is thus defined not by a use case, even around truth seeking, but more by its relationship to Musk and his vow to combat “the woke mind virus.”
All things considered, embracing a specific personality and politics, Claude and Grok seem far better positioned to build emotional loyalty rather than Gemini and Copilot, and especially ChatGPT. Being for a certain kind of person is, relatively speaking, a stickier form of positioning than being for people who already use a software suite, or “everyone” in general.
LLM Positioning Strategy Marketing:
Perplexity Has A Use Case
Perplexity’s positioning, by contrast, is easily the thickest: it focuses on a use case.
Perplexity is primarily named after a technical metric that refers to how well an LLM predicts language, (lower perplexity means a more accurate model). But it obviously also references the state of perplexity and curiosity about something which seems confusing, which everyone experiences. So, while this brand name is also a deep cut reference, like Gemini, Claude, and references the underlying technology like ChatGPT, it leverages a well-known experience like Grok to easily suggest what kinds of problems and solutions it will offer.
Perplexity positioning focuses on it as a research tool that cites in-line sources for claims rather than centering a specific personality or set of political values. I am not suggesting Perplexity is somehow value-free compared to Claude or Grok. In-line-citation-first design is its own set of commitments. But this focus does create a clear use case, which other LLMs lack.
This means Perplexity isn’t for everyone. If you are willing to rely on an LLM that is not optimized to show its work, you are implicitly saying the provenance of your information doesn’t matter much to you. I get that for low stakes tasks (e.g. recipes for chicken gumbo). But ignoring sources undermines the habits that define careful, intellectually-responsible thinking. One of the primary aims of any formal education is to combat this kind of intellectual shortcut.
Simply put, Perplexity is for those who want an LLM with better research habits than a precocious child. It gives numbered sources for each claim as it states them, which reminds you to check them, while making it easy to do so. It still hallucinates and misattributes. Perplexity can still cite sources that only partially support a claim. It doesn’t and can’t remove your responsibility to be skeptical, it just makes it relatively trivial to double check its sources.
Yes, other LLMs can cite sources, but tools like ChatGPT, Gemini, Copilot, Claude, and Grok require you to request citations, or manually switch into a special research mode. They do not default into a “show your work” mode. Instead, they invite you to be intellectually lazy, habitually trust the LLM, and ignore effortful instincts from formal education you’ve received.
Yes, you can force most LLMs to behave more like Perplexity if you are persistent enough, but in practice that is tedious, annoying, and unreliable. This irritating outcome is not surprising. Most LLMs have neglected to pick a specific product niche to begin with, while Perplexity went all-in on being a research tool. In the future, this can easily change though; other LLMs can be designed as more reliable research tools too. But in the meantime, Perplexity is the best in class option for users who care about verifiable sources, and its building loyalty.
LLM Positioning Strategy Marketing:
Why Thick Positioning Wins
In the short term, I can see why the relatively “thin” positioning of ChatGPT, Gemini, Copilot, Claude and Grok has proven more popular than Perplexity “thick” positioning. It is easier to acquire investor money with promises of a genie that can grant infinite wishes. And if you dominate the market by promising everything, you can try to pivot to specifics later.
In the long run, “thin” positioning is a liability. Partly because, if I cannot specifically tell where your tool excels, I cannot build habits around it, and I will abandon it when something purpose-built comes along. At the same time, investors want profits, not promises. That requires consumers willing to pay for specific use cases, sold on specific positioning. The bill on AI tools will come due, and “thin” positioning is not going to create users happy to foot it.
Eventually, to be profitable Perplexity competitors will also need to:
- pick a job
- define what counts as “good” in concrete behavioral terms
- make those commitments visible in the product itself
- position the product to sell that commitment
Insight to Action is a premier positioning strategy marketing firm, working in many industries like tech and food and beverage. Subscribe to our newsletter for regular insights. Contact us to start a conversation about your brand.
