39.9% of Search Patent Families Published in 2026 Mention Generative AI, Up From 0.5% in 2022

Summary

Picture a homeowner asking an AI assistant who repairs tankless water heaters nearby. The assistant runs a search, then writes a short answer that names two or three businesses. That pairing of search with a writing model is new, and the patent record shows how fast it became common in filings.

We looked at 26,486 search-related patent families published since 2018 and checked which ones also describe generative AI or large language models. In 2022, only 12 of 2,488 did (0.5%). In 2026 so far, 794 of 1,988 do (39.9%).

The total across the period is 2,269 patent families. More than half of those published in 2026 describe retrieval-augmented generation, the method where a model looks things up before it writes. A patent shows what was filed, not what is live in any product.

What we measured

We pulled every publication in Google Patents Public Data (patents-public-data.patents.publications) published from January 1, 2018 to September 17, 2026 that carries at least one CPC code in G06F16 (information retrieval), G06F40 (natural language processing) or G06N (AI and machine learning) and has an English abstract. That is 2,170,848 publications in 1,214,491 patent families. For each publication we kept the English title, the English abstract and up to the first 20,000 characters of the English claims. English claims exist for 488,406 publications, mostly US, European and international (WO) filings, so text matching is deeper for those offices than for Chinese, Korean or Japanese filings, which we matched on title and abstract.

A family counts as search-related when its English text uses terms such as search engine, search query, search results, web search, information retrieval or retrieval-augmented generation. It counts as generative AI when the text names large language models, LLMs, generative AI or generative models, GPT, foundation models or pre-trained language models. We counted each patent family once and used the earliest publication date in our window for the year. Priority year is the earliest filing date in the family.

Generative AI went from rare to routine in search patents

We split every search-related family by the year it first published and checked how many also describe generative models.

The share stayed under 1% through 2022, then jumped to 2.3% in 2023, 14.1% in 2024, 30.1% in 2025 and 39.9% in 2026 to date. Publication follows filing by roughly 18 months, so most of this growth reflects filings made in 2023 and 2024.

Retrieval-augmented generation now leads the filings

Retrieval-augmented generation, often shortened to RAG, means the model retrieves documents first and writes its answer from them. That is the pattern behind most AI answers that cite web pages.

No family in the set used RAG language before 2024. By 2026, a majority do. For a business, if products follow these filings, more AI answers will be built from pages the system fetched, which puts more weight on whether your pages can be found and read.

Most of these filings are about querying text

Patent offices tag every filing with CPC classification codes. A family can carry many.

The top class covers search over free text such as documents and web pages. Web retrieval itself (G06F16/95) appears in 337 families, or 14.9%.

Filing activity is highest for 2024 so far, and 2025 is still arriving

Counting by earliest filing date shows when the work started, though recent years are incomplete because of the publication delay.

The 627 families first filed in 2025 are a floor, not a final count. Most 2025 filings will not be public until 2026 and 2027.

What this means

When nearly 40% of new search patents describe a model that writes the answer, it is safe to plan for AI answers being a lasting part of how customers find businesses. The patents point to systems that fetch documents and then summarize them, so being one of the documents they fetch is the goal. Getting a business into that fetched set is the core job of an AI search optimization agency.

Practical steps: keep key facts about your services, prices and locations in plain text on crawlable pages, answer the questions customers actually ask, and make sure nothing blocks reputable crawlers. If you want to see whether AI assistants mention your business today, a free AI visibility audit gives you a starting measure of AI brand visibility.

Limits

A patent shows what was filed, not what is live in any product.

Patents usually publish about 18 months after the first filing, so families first filed in 2025 and 2026 are heavily undercounted, and 2026 publication counts cover January 1 to September 17, 2026 only.

Matching is keyword based. Some families use the terms in passing, and some relevant families use other words and are missed.

English claims are mostly available for US, European and WO filings, so those offices are matched more deeply than offices with English abstracts only. Chinese filings rely on English abstracts provided in the dataset.

The universe is limited to families tagged G06F16, G06F40 or G06N. Relevant filings classified only elsewhere are not counted.

Frequently asked questions

Related studies

• Patents behind AI searchQuery Fan-Out Patents: 44.1% Now Use Generative AI
• Patents behind AI searchFreshness Shows Up in Only 3.6% of AI Ranking Patents
• Patents behind AI searchAI Search Patents: Top 10 Filers Hold Just 17.9%
• Patents behind AI searchOnly 3.9% of AI Search Patents Describe Citing Sources