Meet the AI Platforms10 min read

AI Search Engines

The 10 blue links are dying — AI now reads the internet and answers you directly
scope:Foundationaldifficulty:Beginner

The Death of 10 Blue Links

For over 20 years, searching the internet looked the same. You typed a query. You got a page of 10 blue links. You clicked one, read through ads and cookie banners, maybe found what you wanted, went back, clicked another link, and repeated until you gave up or got lucky.

That era is ending.

In 2023-2025, a new kind of search engine emerged — one powered by LLMs (Large Language Models). Instead of giving you a list of web pages, these AI search engines read the web pages for you, synthesize the information, and give you a direct answer. With sources cited, so you can verify.

It's like the difference between a librarian who points you to a shelf of books... and a librarian who reads all the books, writes you a summary, and tells you exactly which page each fact came from.

The Major Players

Perplexity AI

Perplexity is the poster child of AI search. Launched in 2023, it quickly gained millions of users by offering something Google didn't: direct answers with citations.

  • You ask a question in natural language
  • Perplexity searches the web in real time
  • It reads the top results and synthesizes an answer
  • Every claim gets a numbered citation you can click to verify
  • You can ask follow-up questions in a conversation

It's like having a research assistant who reads 20 articles in 3 seconds and gives you a bibliography.

Google AI Overviews

Google saw Perplexity gaining traction and responded with AI Overviews (originally called "SGE" — Search Generative Experience). Now when you search on Google, an AI-written summary often appears above the traditional blue links.

The catch? Publishers are furious. If Google answers the question directly, fewer people click through to their websites. Traffic drops. Ad revenue falls. This tension between AI answers and the open web is one of the biggest debates in tech right now.

Bing Copilot (Microsoft)

Microsoft integrated GPT-4 directly into Bing in early 2023 — the first major search engine to go full AI. Bing Copilot combines traditional search results with AI-generated answers in a chat interface. You can ask follow-ups, generate images, and even have it compose emails — all within the search experience.

How AI Search Actually Works

Under the hood, AI search engines follow a consistent pattern:

  1. Query understanding — The LLM interprets your question. "What's the best way to learn Python in 2025?" is understood as needing current resources, beginner-friendly options, and comparisons.
  2. Web retrieval — The system searches the web (like a traditional search engine) and pulls back relevant pages.
  3. Reading and extraction — The LLM reads the content of those pages — articles, forums, docs — and extracts relevant facts.
  4. Synthesis — The LLM combines information from multiple sources into a coherent, original answer.
  5. Citation — Each claim is linked back to its source, so you can verify the information.

This is essentially RAG (Retrieval-Augmented Generation) at scale — a pattern we explore in detail in the RAG article.

The Citation Problem

The biggest criticism of AI search is accuracy. AI search engines can:

  • Hallucinate — Make up facts that aren't in any of the sources
  • Misattribute — Cite a source that doesn't actually say what the AI claims
  • Cherry-pick — Present one perspective as fact when the topic is debated
  • Become stale — Use outdated information from cached pages

This is why citations matter so much. Perplexity and Google both include source links — but you should still click through and verify for anything important.

Using Perplexity's API in Python

from openai import OpenAI


# Perplexity uses an OpenAI-compatible API!
client = OpenAI(
    api_key="pplx-your-api-key",
    base_url="https://api.perplexity.ai"
)


# Ask a question — Perplexity searches the web in real time
response = client.chat.completions.create(
    model="sonar",  # Perplexity's search model
    messages=[
        {"role": "system", "content": "Be precise. Cite sources."},
        {"role": "user", "content": "What are the top 3 Python web frameworks in 2025?"}
    ]
)


print(response.choices[0].message.content)


# The response includes inline citations like [1], [2], [3]
# that link back to the web sources it read
Output
Based on current developer surveys and GitHub activity:

1. **Django** [1] — Full-featured framework, great for large apps.
   Used by Instagram, Pinterest, and Mozilla.
2. **FastAPI** [2] — Modern async framework, fastest-growing.
   Built-in OpenAPI docs and type validation.
3. **Flask** [3] — Lightweight and flexible, ideal for microservices
   and small projects.

Sources:
[1] https://survey.stackoverflow.co/2025
[2] https://fastapi.tiangolo.com
[3] https://flask.palletsprojects.com
Note: Will AI search kill websites? This is a real concern. If AI search answers questions directly, people have less reason to visit the original websites. Publishers lose traffic and ad revenue. Some news sites have already started blocking AI crawlers. The future might involve revenue sharing — where AI search engines pay publishers whose content they synthesize. This tension is still unresolved.

When to Use AI Search vs. Traditional Search

AI search isn't always better. Here's a guide:

  • Use AI search for: Research questions, comparisons, "explain this concept," multi-step questions, summarizing a topic, getting up to speed quickly.
  • Use traditional search for: Navigating to a specific website, finding a specific document or page, shopping, local businesses, real-time results (sports scores, stock prices).

The best approach? Use both. Start with AI search for understanding, then use traditional search to find specific resources, verify claims, and go deeper.

The Future of Search

We're in the middle of the biggest shift in search since Google replaced Yahoo. In the next few years, expect:

  • Multimodal search — Search with images, voice, and video, not just text
  • Personalized AI search — Models that learn your preferences over time
  • Agentic search — AI that doesn't just answer questions but takes actions (books flights, fills out forms, compares prices across stores)
  • Source verification — Better tools to detect AI hallucinations and verify claims
Challenge

Quick check

What is the core difference between traditional search and AI search?

Continue reading