# AI search

> Source: https://aiwiki.ai/wiki/ai_search
> Updated: 2026-06-23
> Categories: Artificial Intelligence, Information Retrieval, Natural Language Processing
> From AI Wiki (https://aiwiki.ai), a free encyclopedia of artificial intelligence. Quote with attribution.

**AI search** (also called **AI-powered search**, **generative search**, or an **answer engine**) is a class of search engine and search feature that uses [large language models](/wiki/large_language_model) (LLMs) and [generative AI](/wiki/generative_ai) to produce a direct, synthesized answer to a query rather than returning a ranked list of links. Instead of matching keywords to web pages, an AI search system interprets the intent behind a query, retrieves relevant information from multiple sources, and generates a coherent, cited response that aims to answer the question directly.

The category has grown rapidly since 2023, led by [Perplexity AI](/wiki/perplexity_ai), [ChatGPT](/wiki/chatgpt) search, Google AI Overviews, and Microsoft Copilot (formerly Bing Chat). The scale is already large: Google reported that AI Overviews reached 2 billion monthly users by July 2025, up from 1.5 billion in May 2025, across more than 200 countries and 40 languages [1], and ChatGPT passed 800 million weekly active users in October 2025 [2]. A Pew Research Center analysis of real browsing data found that roughly 18% of all Google searches in March 2025 produced an AI-generated summary, and that users who saw such a summary clicked a traditional link only 8% of the time versus 15% when no summary appeared, a shift that sits at the center of debates over the future of the open web [3].

## How does AI search work?

AI search systems combine several technical components to transform a user query into a generated answer. The process closely resembles [retrieval-augmented generation](/wiki/retrieval_augmented_generation) (RAG) applied to the open web. The 2023 Princeton paper that named the optimization discipline around these systems described the underlying mechanism plainly: "Generative Engines typically satisfy queries by synthesizing information from multiple sources and summarizing them using LLMs" [4].

### Query understanding

When a user enters a query, the system first analyzes it to understand the intent, identify key entities, and determine what type of information is needed. This goes beyond keyword matching. An AI search system can recognize that "best laptop for video editing under $1500" requires product recommendations filtered by use case and price, not just pages containing those words.

Advanced query understanding may include:

- **Query classification:** Determining whether the query is informational, navigational, transactional, or conversational.
- **Entity recognition:** Identifying specific people, places, products, or concepts mentioned in the query.
- **Query expansion:** Generating related sub-queries to cover different aspects of the user's information need.
- **Temporal understanding:** Recognizing when the user wants current information ("latest") versus historical information.

### Web retrieval

After understanding the query, the system retrieves relevant content from the web. Most AI search products use a combination of methods:

- **Web index search:** Searching a pre-built index of web pages, similar to traditional search engines. [Perplexity](/wiki/perplexity), for example, maintains its own web index in addition to using external search APIs.
- **Real-time web crawling:** Some systems crawl specific pages in real time to ensure the most current information.
- **API-based retrieval:** Querying news feeds, knowledge bases, or specialized databases for structured information.

The retrieval step typically returns multiple source pages, which are then processed and ranked for relevance to the query [5].

### Answer generation

The retrieved content is fed into an LLM along with the original query. The model synthesizes information from multiple sources into a coherent, readable answer. This is the generative step that distinguishes AI search from traditional search: rather than presenting a list of links and leaving the user to piece together an answer, the system does the synthesis work.

The quality of answer generation depends on several factors: the relevance of retrieved sources, the model's ability to synthesize conflicting information, the handling of ambiguous or subjective queries, and the system's capacity to distinguish reliable sources from unreliable ones.

### Citation and attribution

A defining feature of AI search (distinguishing it from general chatbot interactions) is inline citation. AI search products typically include numbered references linking specific claims to their source web pages. This allows users to verify information and provides attribution to the original publishers.

Different products implement citation differently. Perplexity provides numbered inline citations linked to source URLs. Google AI Overviews link to the source pages that informed each statement. ChatGPT search includes source links but has been observed to cite differently ranked pages than traditional search results [6].

| Component | Traditional search | AI search |
|---|---|---|
| Query processing | Keyword matching, basic intent classification | Deep semantic understanding, entity recognition, query expansion |
| Retrieval | Index lookup returning ranked list of pages | Multi-source retrieval optimized for answer generation |
| Result presentation | List of 10 blue links with snippets | Generated prose answer with inline citations |
| User effort | User must click through links and synthesize information | System synthesizes information; user reads the answer |
| Interaction model | One query, one result page | Conversational; follow-up questions supported |

## What are the major AI search products?

The AI search landscape has evolved rapidly, with several distinct products competing for users.

### Perplexity AI

[Perplexity AI](/wiki/perplexity_ai), founded in August 2022 by Aravind Srinivas, Denis Yarats, Johnny Ho, and Andy Konwinski, was the first major product built entirely around the AI answer engine concept. Rather than displaying links, Perplexity generates a researched answer to each query with numbered inline citations.

By mid-2025 Perplexity was handling roughly 780 million queries per month and reported about 45 million active users at that stage [7]. The company's valuation rose sharply during 2025: it reached $18 billion in a July 2025 round backed by Nvidia and SoftBank's Vision Fund 2 (an extension of an earlier round that had valued it at $14 billion), then $20 billion in a September 2025 raise of about $200 million [7][8]. Its annual recurring revenue was reported to be approaching $200 million by late 2025 [8]. Perplexity uses multiple underlying models (including its own fine-tuned models and third-party models from [OpenAI](/wiki/openai) and [Anthropic](/wiki/anthropic)) and maintains its own web index for retrieval.

Perplexity has faced controversies around its relationship with publishers. In 2024, Forbes and Wired accused Perplexity of reproducing or paraphrasing their content without authorization. In July 2024 the company responded by launching the Perplexity Publishers' Program, a revenue-sharing arrangement with early partners including Fortune, Time, Der Spiegel, and The Texas Tribune; it expanded the program in December 2024 to outlets in the UK, Japan, Spain, and Latin America [9].

### ChatGPT search (SearchGPT)

OpenAI launched ChatGPT search on October 31, 2024, integrating real-time web search directly into [ChatGPT](/wiki/chatgpt). It was first previewed as the "SearchGPT" prototype on July 25, 2024, was powered by a fine-tuned version of [GPT-4o](/wiki/gpt-4o), and was rolled out to all logged-in users on February 5, 2025 [10]. The feature lets the model search the web when it determines that current information is needed to answer a query.

ChatGPT's reach gives the search feature enormous distribution. OpenAI CEO Sam Altman said in October 2025 that "more than 800 million people use ChatGPT every week," up from 500 million in March 2025 and 700 million in August 2025 [2]. Research has also shown that ChatGPT search tends to cite pages that rank lower in traditional Google search results (position 21 and beyond) a large share of the time, suggesting it surfaces different content than traditional search [6].

### Google AI Overviews

Google launched AI Overviews (initially called Search Generative Experience, or SGE) at its I/O conference on May 14, 2024, integrating AI-generated summaries directly into Google Search results. When a user's query triggers an AI Overview, a generated answer appears at the top of the results page, above the traditional organic links. Announcing the broader AI strategy, CEO Sundar Pichai said the "bold and responsible approach is fundamental to delivering on our mission and making AI more helpful for everyone" [11].

AI Overviews scaled quickly to 2 billion monthly users by July 2025 [1]. According to analytics firm BrightEdge, the feature appeared in over 11% of Google queries by mid-2025, a 22% year-over-year increase [12]. Google's approach is more conservative than standalone AI search products: AI Overviews supplement rather than replace the traditional link results, and they are not triggered for all query types. Google has also said the feature is driving over 10% more searches for the kinds of queries where it appears [1].

Google's implementation initially faced criticism for accuracy issues, including a widely publicized incident in May 2024 in which AI Overviews suggested adding glue to pizza. Google subsequently tightened its quality controls and reduced the scope of queries that trigger AI Overviews [12].

### Microsoft Copilot (Bing Chat)

Microsoft was the first major search engine to integrate generative AI, launching the new Bing with built-in chat on February 7, 2023 using [OpenAI](/wiki/openai)'s [GPT-4](/wiki/gpt-4) technology (wrapped in Microsoft's "Prometheus" model). At the launch event, CEO Satya Nadella declared, "It's a new day for search" [13]. The product has been rebranded multiple times, becoming Microsoft Copilot in late 2023.

Copilot holds roughly 14% of the AI chatbot market by share [14]. Its integration with Microsoft's broader product ecosystem (Windows, Office, Edge browser) gives it distribution advantages, though it has not displaced Google as the dominant search destination.

### Other AI search products

| Product | Launch | Key differentiator | Model(s) used |
|---|---|---|---|
| [Perplexity AI](/wiki/perplexity_ai) | 2022 | Purpose-built answer engine; inline citations | Multiple (proprietary + OpenAI, Anthropic) |
| [ChatGPT](/wiki/chatgpt) search | 2024 | Integrated into dominant chatbot; 800M weekly users | GPT-4o and successors |
| Google AI Overviews | 2024 | Integrated into dominant search engine; 2B monthly users | [Gemini](/wiki/gemini) |
| Microsoft Copilot | 2023 | Integrated into Windows/Office ecosystem | GPT-4 and successors |
| You.com | 2022 | Multi-mode interface (Research, Create, Imagine); customizable | Multiple models |
| Brave Search AI | 2023 | Privacy-focused; independent index (not built on Google/Bing) | Proprietary (Brave Leo) |
| Arc Search | 2024 | Mobile-first; "Browse for Me" feature synthesizes pages | Multiple models |
| Exa | 2023 | Developer-focused search API; semantic search for AI applications | Proprietary |
| Kagi | 2023 | Paid, ad-free search with AI summaries; privacy-focused | Multiple models |

## How is AI search affecting traditional search?

### Google's response

Google's position as the dominant search engine (holding over 90% global market share for traditional search) is being challenged for the first time in decades by AI search alternatives. Google has responded aggressively with AI Overviews, AI Mode (a conversational search interface that surpassed 100 million monthly active users by July 2025), and deep integration of [Gemini](/wiki/gemini) across its products [1].

Data from BrightEdge shows that total Google search impressions rose by over 49% in the first year of AI Overviews, suggesting that AI features may be driving more engagement, even as click-through rates to websites fell by nearly 30% over the same period [12]. In other words, users are seeing more results but clicking through to external sites less often.

Google's AI chatbot, Gemini, has also grown into a strong competitor, reaching roughly 450 million monthly active users for the Gemini app by July 2025 [1]. This positions Google as a leading player in the AI search space even as its traditional search faces disruption.

### Market dynamics

Despite the rapid growth of AI search products, the actual share of total web referral traffic from AI search remains small. AI search platforms collectively account for a low single-digit percentage of total referral traffic from search, and Google still dominates overall search activity by a wide margin [15]. The disruption is happening at the margins, primarily affecting informational queries where users previously needed to visit multiple pages to find an answer.

The search market is fragmenting rather than flipping. Traditional search remains dominant for navigational queries (finding a specific website), transactional queries (shopping, booking), and local search. AI search is strongest for research-oriented, multi-faceted informational queries where synthesis adds the most value.

## Is AI search hurting publishers?

The rise of AI search has created significant anxiety among web publishers, whose business models depend on search-driven traffic.

### Traffic decline

Google search traffic to publishers fell sharply in 2025. Using Chartbeat data, Press Gazette reported that global publisher traffic from Google dropped by roughly a third in the year to November 2025, with the decline steeper in the United States (down about 38%) than in Europe (down about 17%) [16]. Publishers specializing in utility content (weather, TV guides, how-to articles, health information) were among the hardest hit, as these are precisely the types of queries that AI Overviews address directly.

The clearest evidence of the mechanism comes from the Pew Research Center, which analyzed the actual browsing of 900 U.S. adults across 68,879 Google searches in March 2025. Pew found that when an AI summary was present, users clicked a traditional search result link only 8% of the time, compared with 15% when no summary appeared. Only about 1% of users clicked a link inside the AI summary itself, and browsing sessions were more likely to end immediately after a summary appeared (26%) than without one (16%) [3]. This quantifies the broader trend toward "zero-click searches," in which users get an answer on the results page without visiting any source.

### The attribution problem

While AI search products include citations, the traffic value of these citations is debated. Being cited in an AI-generated answer is not equivalent to receiving a direct search click. Pew's finding that only about 1% of AI-summary impressions lead to a click on a summary link underscores how little referral traffic citations may generate [3]. Users may read the synthesized answer and move on without ever visiting the cited source.

This has led to a growing "attribution gap" in which publishers' content is used to train models and generate answers, but the economic value does not flow back proportionally. Several major publishers, including The New York Times, which sued OpenAI and Microsoft on December 27, 2023, have filed lawsuits related to AI training data usage, though these cases focus primarily on model training rather than real-time search retrieval [17].

### Publisher response

Publishers are adapting in several ways:

- **AI licensing deals:** Many publishers are negotiating content licensing agreements with AI companies. OpenAI has signed deals with publishers including The Associated Press, Axel Springer, Le Monde, and others. A large share of surveyed publishers expect AI licensing to provide at least some revenue in the coming years [16].
- **Generative engine optimization (GEO/AEO):** A new discipline focused on making content more likely to be cited by AI search products has emerged. Sometimes also called answer engine optimization (AEO), it parallels traditional [SEO](/wiki/search_engine_optimization) but targets AI retrieval and synthesis systems. See [generative engine optimization](/wiki/generative_engine_optimization) for the methods involved [4].
- **Content differentiation:** Publishers are investing more in original reporting, analysis, and first-person perspectives that AI search cannot easily replicate from synthesized sources.
- **Robots.txt and opt-out:** Some publishers have used robots.txt directives to block AI crawlers, though this is a blunt instrument that may reduce both AI search visibility and traditional search visibility.

### The revenue challenge

| Revenue model | Traditional search era | AI search era |
|---|---|---|
| Display advertising | Driven by page views from search traffic | Reduced as zero-click searches increase |
| Affiliate links | Users click through and convert | AI answers may bypass affiliate content |
| Subscription | Search drives awareness and trial | AI summaries may reduce perceived need to subscribe |
| AI licensing | Not applicable | New revenue stream; terms vary widely |

## What is generative engine optimization?

[Generative engine optimization](/wiki/generative_engine_optimization) (GEO), sometimes called answer engine optimization (AEO), is the practice of structuring and writing content so that it is more likely to be surfaced and cited inside AI-generated answers. The term was coined in a paper titled "GEO: Generative Engine Optimization," first posted to arXiv on November 16, 2023 by Pranjal Aggarwal, Vishvak Murahari, Tanmay Rajpurohit, Ashwin Kalyan, Karthik Narasimhan, and Ameet Deshpande, and published at the ACM SIGKDD conference (KDD 2024) [4].

The authors framed GEO as "the first novel paradigm to aid content creators in improving their content visibility in generative engine responses," and reported that their methods could "boost visibility by up to 40% in generative engine responses" when measured on their GEO-bench benchmark [4]. Effective tactics identified in the literature include adding statistics and quotations, citing sources, and using authoritative, clearly structured language, which contrasts with traditional SEO's emphasis on keywords and backlinks. GEO has since grown into a commercial discipline alongside the traditional [search engine optimization](/wiki/search_engine_optimization) industry, with agencies and tools dedicated to measuring and improving how often a brand or page is cited by AI search products.

## How does AI search compare to traditional search?

The following table highlights the key structural differences between traditional keyword-based search and AI-powered search.

| Dimension | Traditional search (Google, Bing) | AI search (Perplexity, ChatGPT search) |
|---|---|---|
| Result format | Ranked list of web page links | Generated prose answer with citations |
| Speed | Near-instant | 2-10 seconds for answer generation |
| Accuracy | Depends on source quality and user discernment | Depends on retrieval quality and model capabilities; risk of [hallucination](/wiki/hallucination) |
| Freshness | Index updated continuously | Depends on retrieval; some products search in real time |
| Transparency | User sees all sources and can evaluate them | User sees a synthesized answer; must trust the synthesis |
| Advertising | Dominant revenue model (Google earns ~$250B/year from ads) | Still evolving; Perplexity launched sponsored results in 2024 |
| Multi-turn interaction | Limited (related searches) | Full conversational follow-ups |
| Bias | Results influenced by SEO, domain authority, ads | Results influenced by model training, retrieval algorithm, citation patterns |

## Technical architecture

AI search products share a common high-level architecture, though implementations vary.

### Retrieval layer

The retrieval layer is responsible for finding relevant web content. Most AI search products use a combination of:

1. **Traditional search index:** A web-scale inverted index (like those used by Google or Bing) for fast keyword-based retrieval.
2. **[Semantic search](/wiki/semantic_search):** [Embedding](/wiki/word_embedding)-based retrieval that matches queries to documents based on meaning rather than exact keywords.
3. **Hybrid search:** Combining keyword and semantic search for maximum recall.

Some products build their own search infrastructure (Google, Brave), while others leverage existing search APIs (Perplexity uses its own index plus external APIs; ChatGPT search has used Bing's API) [18].

### Generation layer

The generation layer uses an LLM to produce the answer. Key design decisions include:

- **Model selection:** Which LLM powers the generation. Perplexity uses multiple models; Google uses Gemini; ChatGPT search uses GPT-4o.
- **[Grounding](/wiki/grounding):** Techniques to ensure the generated answer is faithful to retrieved sources and does not hallucinate.
- **Citation injection:** Mechanisms for the model to attribute specific claims to specific sources.
- **Safety filtering:** Content moderation to prevent harmful or misleading outputs.

### Ranking and quality

AI search systems must decide which sources to prioritize. This involves:

- **Source reliability scoring:** Assessing the trustworthiness and authority of retrieved pages.
- **Recency weighting:** Prioritizing newer content for time-sensitive queries.
- **Diversity:** Ensuring the answer draws from multiple perspectives rather than a single source.
- **Conflict resolution:** Handling cases where retrieved sources disagree with each other.

## Monetization and business models

The business model for AI search is still evolving and represents one of the biggest open questions in the industry.

### Subscription revenue

Perplexity offers a Pro subscription ($20/month) with access to more powerful models, more queries, and advanced features. ChatGPT Plus ($20/month) and ChatGPT Pro ($200/month) include search capabilities. Google's AI features are available for free within Google Search, while premium features exist within Google One AI Premium.

### Advertising

Perplexity introduced sponsored results in late 2024, displaying ads alongside AI-generated answers. Google continues to show ads alongside AI Overviews, though the placement and format differ from traditional search ads. The advertising model for AI search is complicated because the synthesized answer format provides fewer natural insertion points for ads compared to a traditional list of links.

### Licensing

AI search companies are increasingly paying publishers for content access. These deals provide revenue to publishers while securing legal protection and content quality for the search products. The terms and amounts of these deals vary widely and are often confidential.

## Challenges

### Hallucination and accuracy

AI search systems can generate plausible but incorrect answers, a problem known as [hallucination](/wiki/hallucination). While citation and grounding techniques reduce this risk, they do not eliminate it. Users may place excessive trust in AI-generated answers because of their confident, authoritative presentation.

### Information echo chambers

By synthesizing a single answer from multiple sources, AI search may reduce users' exposure to diverse perspectives. Traditional search at least presents multiple links, allowing users to encounter different viewpoints. AI search risks creating a "single answer" paradigm that obscures disagreement and nuance.

### Sustainability of the publisher ecosystem

If AI search significantly reduces traffic to publishers, the economic foundation that supports web content creation could erode. This creates a potential paradox: AI search depends on high-quality web content for its answers, but may undermine the business models that produce that content.

### Cost

Generating AI answers is significantly more expensive than serving traditional search results. Each AI search query requires LLM inference, which costs orders of magnitude more than a traditional index lookup. At Google's scale (processing many billions of searches per day), even small per-query cost increases represent large sums annually.

### Legal and copyright concerns

The legal framework around AI search and content usage is still developing. Multiple lawsuits are pending regarding whether AI search products' retrieval and display of publisher content constitutes fair use. The outcomes of these cases will significantly shape the industry's trajectory [17].

## Current state (2025-2026)

As of early 2026, AI search is firmly established as a category but has not yet displaced traditional search. Google remains dominant in overall search, though its market share has faced slight pressure for the first time in years. ChatGPT and Perplexity have established significant user bases but collectively account for a small fraction of total search referral traffic.

The industry is converging on a hybrid model in which AI-generated answers and traditional links coexist. Google's approach of adding AI Overviews on top of existing search results exemplifies this hybrid model. Pure AI answer engines like Perplexity are also adding more traditional features (link results, related searches) to complement their generated answers.

Key trends to watch include the evolution of advertising models for AI search, the outcome of publisher lawsuits and licensing negotiations, the maturation of [generative engine optimization](/wiki/generative_engine_optimization) as a discipline, and whether AI search will achieve meaningful referral traffic numbers or remain a complement to traditional search. The relationship between AI search providers and content publishers remains the central tension defining the category's future.

## See also

- [Perplexity AI](/wiki/perplexity_ai)
- [ChatGPT](/wiki/chatgpt)
- [Retrieval-augmented generation](/wiki/retrieval_augmented_generation)
- [Generative engine optimization](/wiki/generative_engine_optimization)
- [Large language model](/wiki/large_language_model)
- [Generative AI](/wiki/generative_ai)
- [Search engine optimization](/wiki/search_engine_optimization)
- [Hallucination (artificial intelligence)](/wiki/hallucination)

## References

[1] TechCrunch. "Google's AI Overviews have 2B monthly users, AI Mode 100M in the US and India." July 23, 2025. https://techcrunch.com/2025/07/23/googles-ai-overviews-have-2b-monthly-users-ai-mode-100m-in-the-us-and-india/

[2] TechCrunch. "Sam Altman says ChatGPT has hit 800M weekly active users." October 6, 2025. https://techcrunch.com/2025/10/06/sam-altman-says-chatgpt-has-hit-800m-weekly-active-users/

[3] Pew Research Center. "Google users are less likely to click on links when an AI summary appears in the results." July 22, 2025. https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/

[4] Aggarwal, P., Murahari, V., Rajpurohit, T., Kalyan, A., Narasimhan, K., Deshpande, A. "GEO: Generative Engine Optimization." arXiv:2311.09735, November 16, 2023 (KDD 2024). https://arxiv.org/abs/2311.09735

[5] Google Cloud. "What is Retrieval-Augmented Generation (RAG)?" https://cloud.google.com/use-cases/retrieval-augmented-generation

[6] Averi. "ChatGPT vs. Perplexity vs. Google AI Mode: The B2B SaaS Citation [Benchmarks](/wiki/benchmarks) Report." 2026. https://www.averi.ai/how-to/chatgpt-vs.-perplexity-vs.-google-ai-mode-the-b2b-saas-citation-benchmarks-report-(2026)

[7] Bloomberg. "AI Startup Perplexity Valued at $18 Billion With New Funding." July 17, 2025. https://www.bloomberg.com/news/articles/2025-07-17/ai-startup-perplexity-valued-at-18-billion-with-new-funding

[8] TechCrunch. "Perplexity reportedly raised $200M at $20B valuation." September 10, 2025. https://techcrunch.com/2025/09/10/perplexity-reportedly-raised-200m-at-20b-valuation/

[9] Perplexity. "Introducing the Perplexity Publishers' Program." July 30, 2024. https://www.perplexity.ai/hub/blog/introducing-the-perplexity-publishers-program

[10] OpenAI. "Introducing ChatGPT search." October 31, 2024. https://openai.com/index/introducing-chatgpt-search/

[11] Google. "Google I/O 2024: An I/O for a new generation (Sundar Pichai keynote)." May 14, 2024. https://blog.google/inside-google/message-ceo/google-io-2024-keynote-sundar-pichai/

[12] BrightEdge. "One Year Into Google AI Overviews, BrightEdge Data Reveals Google Search Usage Increases by 49%." May 2025. https://www.brightedge.com/news/press-releases/one-year-google-ai-overviews-brightedge-data-reveals-google-search-usage

[13] TechCrunch. "Microsoft launches the new Bing, with ChatGPT built in." February 7, 2023. https://techcrunch.com/2023/02/07/microsoft-launches-the-new-bing-with-chatgpt-built-in/

[14] First Page Sage. "Top Generative AI Chatbots by Market Share, 2026." https://firstpagesage.com/reports/top-generative-ai-chatbots/

[15] Press Gazette. "Publishers say Google search traffic in 'managed decline', not dead." 2026. https://pressgazette.co.uk/publishers/search-isnt-dead-its-fragmenting-how-to-manage-google-traffic-decline/

[16] Press Gazette. "Global publisher Google traffic dropped by a third in 2025." 2026. https://pressgazette.co.uk/media-audience-and-business-data/google-traffic-down-2025-trends-report-2026/

[17] The New York Times. "The Times Sues OpenAI and Microsoft Over A.I. Use of Copyrighted Work." December 27, 2023. https://www.nytimes.com/2023/12/27/business/media/new-york-times-open-ai-microsoft-lawsuit.html

[18] PragoMedia. "New Data Confirms Top Google Rankings Matter for ChatGPT, Perplexity & AI Search." March 2026. https://www.pragomedia.com/2026/03/09/new-data-confirms-top-google-rankings-matter-for-chatgpt-perplexity-ai-search/