5 min readMar 27, 2024

**Google vs. Perplexity AI: A Product Manager’s Perspective**

In the first article of my series of reviewing AI technology trends, as a product manager I would try navigating the evolving landscape of web search technologies, and the clash between Google and Perplexity AI which is nothing short of a strategic chess match. Let’s dissect their approaches, strengths, and how the veteran Google is gearing up to take on this young challenger, which has started to look formidable esp. after titans like Jeff Bezos, chipmaker Nvidia, Nat Friedman, Elad Gill, etc have invested in the company.

**Understanding the Battlefront**

A) Google’s Algorithmic Citadel

1. Algorithmic Heritage

○ Google’s legacy lies in its algorithmic prowess.

○ The PageRank algorithm, developed by Sergey and Larry ranks pages in which the keywords searched have appeared. This algorithm akin to a medieval castle, fortified Google’s position as a search giant.

2. Semantic Search and BERT (Bidirectional Encoder Representations from Transformers)

○ Google’s BERT model, like a machine diplomat, understands context and nuances. BERT is primarily designed for contextual comprehension in natural language understanding. It considers both the words that come before and after a given word, to understand the user’s context and making it effective for various NLP tasks.

○ It bridges the gap between user intent and search results yet doesn’t have the contextual understanding of the level of a modern LLM like ChatGPT, on which Perplexity is built. We’ll see the reason later in this article

3. Knowledge Graph: The Web of Alliances

○ Google’s Knowledge Graph acts as a web of alliances.

○ It connects entities, providing context and enhancing search results.

B) Perplexity AI’s Conversational Arsenal
1. Chat-Based Warfare
○ Perplexity AI enters the fray with conversational AI.
○ Users engage in dialogue, and not just queries. User is encouraged to follow each query with next prompt and algorithms take care of the entire series of prompts as a single conversation rather than dealing with them separately, unless specified.
○ It’s like having a trusted advisor rather than a search engine where user can keep digging deeper into the topic.
2. Generative AI and Summarization
○ Perplexity AI wields generative AI like a swift blade.
○ Its contextual understanding of user’s search query (prompts) comes from LLM i.e. ChatGPT. Models, including ChatGPT, aim to provide conversational experiences. They generate responses based on the input prompt and can handle a wide range of contexts, from casual chats to more complex interactions because they consider probability of every next word in a sentence and relate the context of whole sentence together (unlike BERT which considers 3 words at a time) and in the same way of a group of sentences in a query/prompt given by the user.

**Strengths and Vulnerabilities**
C) Perplexity AI’s Hidden Daggers
1. Conversational Efficiency
○ Perplexity AI’s strength lies in its efficiency of contextual understanding, i.e., understanding word after word to figure out the context of sentence and sentence after sentence to figure out context of overall text supplied as a query/prompt.
○ Academics and researchers benefit from Perplexity search as it allows to systematically dig deeper while keeping the context intact.
2. Precision Strikes
○ Have a look below at the prompt supplied to both Google and Perplexity AI and compare the precision. Google’s response looks purely out of context while Perplexity has handled it better.
○ Niche topics may find their champion here.
D) Google’s Fortress of Dominance
1. Infrastructure Monolith
○ Google’s colossal infrastructure ensures reliability.
○ Billions of searches flow seamlessly through its data centers and enforces BERT’s self learning, though algorithm architecture might hamper effectiveness of this self learning
2. Ecosystem Hegemony
○ Google’s services—Gmail, Maps, Drive, Earth, etc. —form an ecosystem.
○ Users are entrenched, reluctant to abandon familiar walls.

  • *Chinks in the Armor**
    E) Perplexity AI’s Achilles' Heel
    1. Language Model Dependency
    ○ Perplexity AI leans heavily on language models which it doesn’t owns.
    ○ Niche queries may pierce its armor probably because of existing potential to improve its web scrapping ability as compared to Google.
    2. User Base Expansion
    ○ Perplexity AI must recruit beyond early adopters.
    ○ Scaling its forces is critical. It would need to balance between monetisation and client acquisition/active users, which is currently more aligned towards the former.
    F) Google’s Vulnerabilities
    1. Privacy Quandaries
    ○ Google’s data collection practices raise eyebrows.
    ○ Users demand transparency and control.
    2. Information Overload and irrelevant search results
    ○ Google’s comprehensive results sometimes overwhelm besides displaying irrelevant results and advertisements (paid searches) that too at the top of the search results.
    ○ Finding the Holy Grail —the most relevant answer—requires better contextual understanding using the underlying algorithm.
    **Strategic Maneuvers**
    G) Google’s Counteroffensive
    1. AI Reinforcements
    ○ Google is heavily investing in AI research through its Deepmind project while trying to master LLMs through Gemini.
    ○ It aims to enhance search quality and user satisfaction.
    ○ BERT’s allies multiply — BERT can be replaced by or aided through the LLM based or deeplearning based algorithms to improve its contextual understanding and impart conversational web search ability.
    2. Misinformation Crusade
    ○ Google prioritizes reliable sources.
    ○ It has fact-checking knights guard against fake news dragons, and have been under scrutiny of regulators time and again
    3. User-Centric Innovations being tried
    ○ Google’s secret weapon: understanding user intent.
    ○ Personalized search experiences are the new excalibur.

**Other Knights at the Round Table**
1. ChatGPT by OpenAI
○ ChatGPT, a noble contender, wields conversational AI.
○ It jousts with both Google and Perplexity AI. Like BERT, however, the ChatGPTs training data is not updated. At the same time, due to lack of a side web scrapping tool unlike Perplexity, ChatGPT couldn’t provide real time search results which have incorporated latest scientific research or data
18. Bing’s Silent March
○ Microsoft’s Bing, bolstered by AI, eyes the throne. Microsoft already owns ChatGPT (they avoid marketing this 😉) may have lost to Google in the past but is in full mood to take on its arch rival this time
○ It’s a dark horse in this epic saga.
In this saga of search supremacy, the battlefield shifts daily. While Google still is largely the favorite search engine, most of us shall have ChatGPT open in one of its tabs. While Google has most trained contextual understanding algorithm, it is challenged in terms of concept it is built upon. However, page ranking and crawling ability of Google are still amongst the best. At the same time Perplexity is bringing state of the art contextual understanding combined with real time web scrapping to display the most relevant search results in real time. It would be interesting to see where and how Bing will position itself after it theoritically owns ChatGPT already. ChatGPT isn’t a search product though, more focussed on developing and constantly improving first line of GenAI infrastructure which can be subsequently used by other products. As product managers, we shall watch, try and understand strategies at the scale of best companies and best minds in the silicon valley, learn and adapt.

Will Perplexity AI’s conversational blade pierce Google’s algorithmic armor? Only time— ecosystem building and leveraging capacity and above all, user loyalty—will tell. 🗡️🔍

PerplexityAI search result for query "who am I"
Google search result for query "who am I"

AI Product Management | AI enthusiast | Ever learning | Currently working as specialist product manager in R&D and Imnovation Lab of a 6000 strong workforce