What is Diffbot?
Diffbot, from Diffbot Technologies Corp. of Menlo Park, California, reads web pages with computer vision and language processing rather than a large language model, and returns them as structured JSON. Its Extract API has fixed page types: Article, Product, Image, Video and Discussion, with Event, List and Job in beta, plus Analyze, which works out the page type for you. No per-site rules are needed.
Around Extract sit Crawl, which spiders a whole site from one seed URL, a Natural Language API that pulls entities, relationships and sentiment out of plain text, a Web Search API, and the Knowledge Graph: over 10 billion entities such as companies, people and articles, built from the public web and queried with Diffbot's own query language, DQL, or matched against your records with Enhance.
Everything is paid in credits. Extracting a page costs 1 credit (2 through a proxy), a search 1, a Natural Language document of up to 10,000 characters 1, and a Knowledge Graph entity 25 to 100.
How much does Diffbot cost?
| Plan | Price | What you get |
|---|---|---|
| Free | Free | 10,000 credits/month, 1 thread, 5 calls/minute; blocked when credits run out |
| Startup | $299/month | 250,000 credits, $0.001/credit overage, 5 calls/second |
| Plus | $899/month | 1,000,000 credits, $0.0009/credit overage, Crawl, 3 seats, 25 calls/second |
| Enterprise | Custom | 100+ active crawls, custom limits and seats |
Free: 10,000 credits/month, 1 thread, 5 calls/minute, no card. Startup $299/month for 250,000 credits ($0.001/credit after). Plus $899/month for 1,000,000 credits ($0.0009/credit after), with Crawl and 3 seats. Enterprise custom. Page extract 1 credit; Knowledge Graph entity 25 credits. Startup tier free for students.
Is Diffbot worth it?
Scraping product prices or news articles from a hundred different sites usually means a hundred different parsers, each one quietly breaking the day a site redesigns.
Diffbot was built to end that chore. You send it a URL and it reads the page the way a person would, using computer vision rather than CSS selectors, and hands back typed fields: a product's price, brand and reviews, or an article's author, date and body text. It is not a large language model, so it does not invent a price that is not on the page. One extraction costs 1 credit.
The Knowledge Graph is the part rivals cannot copy cheaply. Diffbot crawls the public web itself and joins what it finds into more than 10 billion entities. You can ask for every software company in Berlin with more than 200 staff, or send your CRM rows and get them back filled in. That costs 25 credits per entity, so 10,000 company records use the whole 250,000 credits of a $299 Startup month.
The catch: the free plan is a test drive, not a tier you can build on. It allows 5 calls a minute, one thread, 400 Knowledge Graph entities a month and no Crawl, and once the 10,000 credits are gone requests are refused rather than billed. The next step is $299 a month, and whole-site crawling starts at $899. Diffbot also keeps your query history and API calls against your key and uses them to improve the service; its privacy policy, updated 29 August 2025, sets no retention period and does not mention training either way. It warns that calls made over plain http travel unencrypted, so use https.
Our take: Diffbot is the reference library, not the photocopier. If you only need pages turned into Markdown for a model, a token-priced reader does it for less; if you need the same fields from thousands of different sites, or the companies and people behind them, Diffbot is the one doing that work without a model guessing.
Where it earns its place
A data engineer at a price-comparison site tracking 300 retailers
Pull price, availability and specs from product pages without writing a parser per shop
Call the Product endpoint for each URL. At 1 credit a page, 900,000 pages a month fit inside Plus at $899, where 25 calls a second keeps a daily refresh realistic.
A revenue-operations lead at a 60-person B2B software firm
Fill in missing headcount, industry and location for 8,000 CRM accounts
Upload the records to Knowledge Graph Enhance. At 25 credits each, 8,000 accounts use 200,000 credits, inside one $299 Startup month; refreshing an enhanced record later costs 100 credits.
A graduate researcher studying news coverage
Turn thousands of news articles into clean text with author and date
Diffbot for Students grants Startup-tier access for free, so 250,000 Article extractions a month cost nothing. Sentiment and entities come from the Natural Language API at 1 credit per document.
Before you commit
- Setup
- Sign up without a card, copy the token and send a GET with token and url to an Extract endpoint; a Test Drive page works without signing in.
- Integrations
- REST APIs for Extract, Crawl, Natural Language, Knowledge Graph (DQL and Enhance) and Web Search; an Excel add-in and browser extensions.
- Data and privacy
- Diffbot logs query history and API calls against your key and uses usage data to improve the service; storage on Amazon S3 and EC2 in the US; no retention period published; account deletion by email to privacy@diffbot.com.
- Cost at scale
- 1.5 million credits a month on Plus is $899 plus 500,000 × $0.0009, about $1,349; Knowledge Graph work burns credits 25 to 100 times faster than page extraction.
- Limits
- Free 5 calls/minute, 400 Knowledge Graph entities and 500 Natural Language calls a month; Startup 5 calls/second; Plus 25 calls/second, 25 active crawls; 429 when credits run out on Free.
What’s good
- Typed fields from any article or product page with no per-site rules
- Not a language model, so no invented values
- Knowledge Graph of over 10 billion entities from its own crawl
- Plain credit pricing with no contract and monthly cancellation
- Startup tier free for students and academics
What’s not
- Free plan capped at 5 calls a minute and blocked when credits run out
- Jumps from $0 to $299 a month
- Whole-site Crawl only from $899 a month
- Knowledge Graph records cost 25 to 100 credits each
- Privacy policy sets no retention period for query history
Best for
Data and sales-intelligence teams who need consistent fields from many different websites, or company and people records, without maintaining scrapers.
Diffbot alternatives
Diffbot: common questions
- Is Diffbot worth it?
- Diffbot pulls clean fields out of pages with no rules to write and owns a company and people graph nobody else sells by the record, but the jump from free to $299 a month is steep. We rate it 3.9 out of 5.
- How much does Diffbot cost?
- Free: 10,000 credits/month, 1 thread, 5 calls/minute, no card. Startup $299/month for 250,000 credits ($0.001/credit after). Plus $899/month for 1,000,000 credits ($0.0009/credit after), with Crawl and 3 seats. Enterprise custom. Page extract 1 credit; Knowledge Graph entity 25 credits. Startup tier free for students.
- What are the downsides of Diffbot?
- Free plan capped at 5 calls a minute and blocked when credits run out; jumps from $0 to $299 a month; whole-site Crawl only from $899 a month; knowledge Graph records cost 25 to 100 credits each; privacy policy sets no retention period for query history.
- Who is Diffbot best for?
- Data and sales-intelligence teams who need consistent fields from many different websites, or company and people records, without maintaining scrapers.
- How much does Diffbot cost for a team or in production?
- 1.5 million credits a month on Plus is $899 plus 500,000 × $0.0009, about $1,349; Knowledge Graph work burns credits 25 to 100 times faster than page extraction.
- Is Diffbot safe for company data?
- Diffbot logs query history and API calls against your key and uses usage data to improve the service; storage on Amazon S3 and EC2 in the US; no retention period published; account deletion by email to privacy@diffbot.com.
- What are the best Diffbot alternatives?
- The closest alternatives we have reviewed are Apify, Jina Reader and TinyFish.
Answers checked Nov 13, 2026.