Counter-positioning is a business strategy in which a company positions itself in a way that its competitors are unwilling to replicate to avoid cannibalization. A well-known example of counter-positioning is Netflix’s policy not to charge late fees. In 2000, Blockbuster was earning a large portion of its revenue ($800 million) from late fees. When Netflix entered the market, the company began shipping DVDs to customers’ homes. Customers could pay for up to three DVDs at a time, and if they didn’t return them, Netflix simply wouldn’t send the next one on the list. This strategy caused Blockbuster to eliminate late fees in 2004, i.e., Netflix essentially forced Blockbuster to cannibalize its business model in order to survive.
By integrating ChatGPT, Bing is poised to put Google in a similarly difficult situation. In 2021, Google earned $148.95 billion (out of $256.74 billion) from search ads. The more users click on different results and reformulate requests, the more advertisers are willing to pay to appear at the top… the more Google generates revenue.
ChatGPT and other language models (LLMs) threaten this business model. Because LLMs generate a single answer that combines different sources, users may often be satisfied with this one, more exhaustive result. There are exceptions to this, such as when users want to find different opinions or browse results. However, most business-related queries can be fully satisfied with language models (e.g., “give me the best hotel under 200 dollars in Austin TX, by the campus, with a minimum average rating of 4.5 stars out of 5”).
Bing’s counter-positioning not only creates collateral damage to Google, but it also lowers Bing’s operating costs: Bing won’t be paying content creators commissions (at least for now) while delivering great value to customers. No wonder Google’s market valuation plunged $100 billion market cap the minute the company demoed Bard – its own LLM that the company will soon integrate into Google Search. The $100 billion drop has been attributed to Google’s failed demo, but there are reasons to believe the real cause lies in investors’ fears to see Google cannibalizing its business model.
The big question now is whether Google is right to cannibalize its golden goose. On the one hand, Google’s strategy to avoid the Blockbuster experience seems right. ChatGPT gained 100 million users in just 2 months, making it the “fastest-growing consumer internet app ever.” To say that there is an appetite for these LLMs and that users seem willing to switch search engines to access them is an understatement. On the other hand, LLMs reduce the ability of search engines to benefit from learning effects (i.e., the improvement in performance that occurs as a result of practice, repetition, or experience). Search engines with LLMs can factor in whether users generate multiple answers, but they can no longer analyze whether users prefer the third or fourth result on the list. Plus, users might get the impression that search engines are endorsing the results provided by LLMs, even if they are wrong. Integrating LLMs may make search more enjoyable, but it could reduce trust in the product. Finally, LLMs rely on averages, and thus exclude singularities, new discoveries, and information that deviates from the consensus.
We’ll have to wait and see if cannibals have a place in this dynamic market. For now, let’s note that competition may not have been one click away, but it surely was one prompt away.
Citation: Thibault Schrepel, Competition Is One Prompt Away, Network Law Review, Fall 2023.