--- title: SerpApi Rolls Out Markdown Output, Cutting Token Overhead for AI Pipelines by Up to 90% | daily.dev description: SerpApi has added Markdown output to all 100+ of its search-data APIs, offering teams a token‑light alternative to JavaScipt Object Notation (JSON) without new endpoints or altered integrations. canonical: https://daily.dev/program/serpapi-markdown-output/ og:type: article og:url: https://daily.dev/program/serpapi-markdown-output/ og:title: SerpApi Rolls Out Markdown Output, Cutting Token Overhead for AI Pipelines by Up to 90% | daily.dev og:description: SerpApi has added Markdown output to all 100+ of its search-data APIs, offering teams a token‑light alternative to JavaScipt Object Notation (JSON) without new endpoints or altered integrations. og:image: https://media.daily.dev/image/upload/s--PT1gkixR--/f_auto,q_auto/v1/recruiter-landing/Screenshot_2026_09_23_at_21_17_12_2b0bcc01ac?_a=BAMAMiB80 og:site_name: daily.dev og:locale: en_US article:published_time: 2026-09-23 article:modified_time: 2026-09-23T18:27:31.936Z article:author: Veronika Furs twitter:card: summary_large_image twitter:site: @dailydotdev twitter:creator: @dailydotdev twitter:title: SerpApi Rolls Out Markdown Output, Cutting Token Overhead for AI Pipelines by Up to 90% | daily.dev twitter:description: SerpApi has added Markdown output to all 100+ of its search-data APIs, offering teams a token‑light alternative to JavaScipt Object Notation (JSON) without new endpoints or altered integrations. twitter:image: https://media.daily.dev/image/upload/s--PT1gkixR--/f_auto,q_auto/v1/recruiter-landing/Screenshot_2026_09_23_at_21_17_12_2b0bcc01ac?_a=BAMAMiB80 --- > SerpApi has added Markdown output to all 100+ of its search-data APIs, offering teams a token‑light alternative to JavaScipt Object Notation (JSON) without new endpoints or altered integrations. Announced in late August 2026, the feature trims token consumption by roughly half on average; some APIs shed as much as 90 percent. In one real‑world test, a Google search for "coffee" fell by nearly 4x from 24,723 tokens in JSON to 6,435 in Markdown. It dropped to just 1,298 tokens when field filtering was applied. ## Why it matters for AI infrastructure For groups running LLM‑powered agents, retrieval pipelines, or search‑grounded applications, the change is one of format rather than architecture. Markdown output slots into existing API calls, so operations teams can adopt it without rewriting orchestration, caching, or monitoring layers. Because context windows remain a hard constraint, shaving token overhead at the source allows models to ingest more relevant search results per request. The result is lower cost per query and higher effective throughput for AI workloads dependent on live search data. ## What changed, and why SerpApi’s Markdown output converts the same structured search data available in JSON into clean, model‑friendly text. Tracking‑heavy redirect links, favicons, and deeply nested metadata are stripped or flattened; duplicate fields are merged into single representations. "As developers increasingly build AI agents and LLM‑powered applications, the way search data is consumed is changing," said Julien Khaleghy, CEO and founder of SerpApi. "When data is consumed by AI, it doesn’t need many elements necessary for JSON. Markdown has emerged as the dominant language of AI models.” "The real benefit is making better use of an AI model’s context window," Khaleghy added. "Because context is limited, the input should be concise and focused on the most important information. Our Markdown output removes unnecessary data, helping AI models process more useful information with fewer tokens." ## Three operational levers Markdown hits three operational levers at once. These include: * Unit cost per query. Token billing is linear; if your agent makes thousands of tool calls a day, halving the payload size translates directly into lower monthly spend without touching prompts or model choice. * Effective throughput. Context windows are a hard cap. Smaller tool responses mean more search results per request, fewer round trips to gather the same evidence, and less risk of hitting token limits mid‑flow. * Cache and storage footprint. If you cache your tool responses (for RAG, evals, or replay), Markdown’s lighter shape reduces object size in your store and egress when those cached blobs are re‑fed to models. ## Availability and how to use it Markdown output is live across all SerpApi plans, APIs, and supported search engines at no additional cost. Developers can request it in three ways: via an 'output=md' query parameter, a '/search.md' route extension, or an 'Accept: text/markdown' header on existing integrations. No new endpoint is required. Teams already using SerpApi's JSON responses can switch formats by adjusting one parameter or header, then measure token usage before and after to confirm savings for their workflows. ### About SerpApi Founded in 2017, [SerpApi](https://serpapi.com/markdown-output) was the first platform to provide developers and enterprises with real‑time access to structured search results from Google, Bing, YouTube, and other sources. The company now offers more than 100 specialized APIs and recently surpassed 1.5 million activated user accounts. SerpApi’s customers include developers, researchers, Fortune 500 companies, and organizations worldwide, using its APIs for AI development, machine learning, SEO monitoring, cybersecurity intelligence, and e‑commerce analysis. The Markdown launch extends that infrastructure to token‑efficient AI pipelines without requiring new contracts or migration projects. --- *This article was produced by SerpApi through the daily.dev Partner Content program. The daily.dev editorial team had no role in its preparation, and it does not reflect daily.dev's editorial positions.* ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/og-image.png?v=a830cdf1","width":1200,"height":630},"sameAs":["https://twitter.com/dailydotdev","https://www.linkedin.com/company/dailydotdev","https://github.com/dailydotdev","https://www.instagram.com/dailydotdev"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","description":"Free, personalized developer news aggregator. Stay on top of software development news, AI coding tools, and web dev - curated daily from trusted sources.","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}},{"@type":"WebPage","@id":"https://daily.dev/program/serpapi-markdown-output/","url":"https://daily.dev/program/serpapi-markdown-output/","name":"SerpApi Rolls Out Markdown Output, Cutting Token Overhead for AI Pipelines by Up to 90%","description":"SerpApi has added Markdown output to all 100+ of its search-data APIs, offering teams a token‑light alternative to JavaScipt Object Notation (JSON) without new endpoints or altered integrations.","inLanguage":"en-US","isPartOf":{"@id":"https://daily.dev/#website"}},{"@type":"AdvertiserContentArticle","@id":"https://daily.dev/program/serpapi-markdown-output/#article","headline":"SerpApi Rolls Out Markdown Output, Cutting Token Overhead for AI Pipelines by Up to 90%","url":"https://daily.dev/program/serpapi-markdown-output/","datePublished":"2026-09-23","dateModified":"2026-09-23T18:27:31.936Z","isPartOf":{"@id":"https://daily.dev/#website"},"publisher":{"@id":"https://daily.dev/#organization"},"mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/program/serpapi-markdown-output/"},"description":"SerpApi has added Markdown output to all 100+ of its search-data APIs, offering teams a token‑light alternative to JavaScipt Object Notation (JSON) without new endpoints or altered integrations.","image":{"@type":"ImageObject","url":"https://media.daily.dev/image/upload/s--PT1gkixR--/f_auto,q_auto/v1/recruiter-landing/Screenshot_2026_09_23_at_21_17_12_2b0bcc01ac?_a=BAMAMiB80"},"author":{"@type":"Person","name":"Veronika Furs"},"timeRequired":"PT3M","isAccessibleForFree":true,"sponsor":{"@type":"Organization","name":"SerpApi"}}]} ```