1. Definition of Data Scraping
Data scraping is the practice of pulling information out of a source that was built for something other than automated, structured retrieval, and converting it into a form software can work with, such as rows in a spreadsheet or records in a database. Dictionary definitions place data scraping alongside the broader term "scraping," describing it as taking information from an output built for a screen or a document and putting it into an ordered record.1 The unifying idea across every variant of scraping is that the source was not designed as a machine-readable feed, so the tool doing the scraping has to interpret a format meant for something else, typically for a human to look at.
2. Human-Readable and Machine-Readable Inputs
A file or page is human-readable when it is organized primarily so a person can understand it by looking at it: a printed invoice, a rendered web page, a PDF report. A file is machine-readable when software can reliably pull specific values out of it without a person retyping or reformatting anything, because the format follows a predictable, documented structure.2 Scraping specifically targets the first category. When a source is already machine-readable, such as a documented API response, retrieving it is a matter of reading a known format rather than inferring structure from something built for a different purpose, which is why that case is usually not described as scraping at all.
3. Data Scraping versus Parsing
Parsing is the general technique of taking a piece of formatted text and turning it into an addressable structure a program can query: parsing JSON text into an object, parsing HTML into a DOM tree, parsing a CSV file into rows and columns. Parsing does not require the input to be poorly structured or unintended for machines; a program parses an API's JSON response just as much as it parses a scraped page's HTML. Scraping almost always includes a parsing step, since the fetched content has to be turned into something addressable before individual values can be pulled out of it. The difference is one of scope: parsing describes how formatted text becomes structure, while scraping describes the larger activity of retrieving content from a source that was not built to be consumed this way and then, typically, parsing it as part of that process.
4. Data Scraping versus Web Scraping
Web scraping is the specific case of data scraping where the source is a web page. It inherits the general definition of scraping (extracting data from a source not designed as a machine-readable feed) and narrows it to content delivered over HTTP as HTML.1 Every web scraping operation is a data scraping operation, but data scraping is the wider category: it also covers scraping a PDF report, a desktop application's window, or a printed page, none of which are web pages. In practice the two terms are frequently used as near-synonyms, since the web is the most common source people mean when they say "scraping" without qualification, but web scraping is properly a subset rather than an equivalent term.
5. Data Scraping versus Screen Scraping
Screen scraping describes reading whatever is currently rendered on a display and treating that rendered output, rather than any underlying data format, as the source of truth. The term predates the modern web, originating with tools that read text directly off terminal screens meant for human operators. Applied to today's software, screen scraping can mean reading a desktop application's window, or reading a web page's rendered, on-screen state rather than its underlying HTML or DOM. Web scraping usually works one level below that, against a page's HTML or DOM structure rather than its rendered pixels, which is a meaningful distinction: DOM-based extraction can reach elements that are not currently visible on screen, while screen scraping in the strict sense is limited to what is actually displayed at that moment.
6. Scraping versus API Access
An API is an interface a data source's owner builds and documents specifically so other software can request data in a defined, stable format. Scraping is what remains available when no such interface exists, or when the interface that does exist does not cover what is needed. Research on scraping in scientific data integration frames this directly: web services are the standard approach to combining data from different sources, but a meaningful number of databases and tools do not expose a web service at all, and even where one exists it may not cover every data need, which is why scraping techniques continue to see practical use.6 An open textbook on data science draws a similar line by treating web scraping and API-based collection as two separate, named methods available for retrieving the same kind of information, one working from a page's rendered content and the other from a documented endpoint.7 Where a suitable API exists, it is generally more stable than scraping, because it is a contract the provider maintains; where it does not, scraping remains the available option.
7. Scraping versus Data Extraction
Data extraction is the broadest term of the group: it covers retrieving selected information from any source, structured or not, and preparing it for further use. Academic surveys of the field describe web data extraction specifically as a topic studied through varied techniques across a broad range of applications, some purpose-built and some reusing methods from the wider field of information extraction.8 Under this framing, scraping is one method of data extraction, specifically the method used when the source was not designed for structured retrieval. Data extraction also covers cases scraping does not, such as pulling fields directly out of a database export or a structured document format, where there is no need to infer structure from something built for a different purpose.
8. Common Source Types
Sources commonly associated with scraping and data extraction include web pages, PDF documents, scanned or photographed material, spreadsheets and reports distributed as files, the rendered interfaces of desktop or mobile applications, and older systems that only expose a terminal-style text interface. What these share is that none of them were built as a documented, machine-readable feed for the specific data a person now wants out of them, even though several of the formats themselves (a PDF, a spreadsheet) have their own internal structure.
9. Common Output Formats
Regardless of which of these methods produced the data, the end goal is usually the same: structured records in a format other software can consume directly. Common targets include CSV, a plain-text tabular format for spreadsheets and simple pipelines, and JSON, a nested, language-independent format widely used for both API responses and general data interchange.3 Larger extraction jobs often write directly into a database instead of an intermediate file, particularly when the extracted data feeds an ongoing application rather than a one-time analysis.
10. Reliability and Interface Changes
Because scraping and screen scraping work against a source's current presentation rather than a documented contract, they are sensitive to changes in that presentation. A redesigned page, a renamed field on a form, or a new version of an application's interface can silently break extraction logic that depended on the previous layout, even though the underlying information a person would see has not changed. Parsing a documented format and calling a stable API are comparatively more reliable, because both operate against a specification the source's owner maintains and is expected not to break without notice. This difference in reliability is one of the clearest practical reasons to prefer an API or a documented export when one is available, and to treat scraping as the fallback for sources that do not offer one.
11. Authorization, Privacy, and Platform Rules
Whether a specific scraping or extraction activity is permitted depends on the source's terms of service, any technical access controls in place, and the applicable law in the relevant jurisdiction, none of which are uniform across sites or countries. This article does not assert that scraping is generally lawful or generally unlawful; that determination depends on the specific source, the data involved, and the context of use, and readers should evaluate their own situation rather than treat any general statement here as legal guidance. Separately from legality, data obtained through scraping can include personal information, which carries its own privacy considerations regardless of how the data was technically retrieved, and platforms commonly apply rate limits or technical restrictions that constrain scraping activity independent of any legal analysis.
12. Comparison Table
| Typical source | Intended original audience | Level of structure | Common method | Typical output | |
|---|---|---|---|---|---|
| Data scraping | Any output not built as a machine-readable feed | People or a specific application | Low, as found | Automated reading of the rendered or published output | Structured records |
| Web scraping | Web pages | People browsing the site | Low to moderate, HTML markup only | HTTP fetch plus HTML or DOM parsing | Structured records |
| Screen scraping | A currently rendered display or window | The human operator of that display | Low, whatever is on screen | Reading the visible, rendered output directly | Structured records |
| Parsing | Any formatted text or document | Varies, machine or human | Matches the source format | Applying a format-specific grammar to build an addressable structure | An in-memory structure or extracted fields |
| API access | A documented endpoint | Developers and programs | High, defined by the provider | Requests to a documented interface | Structured data in a defined schema, often JSON |
| Data extraction | Any source, structured or not | Varies | Varies by source | Whichever method fits the source, including scraping and parsing | Structured records for further use |
Limitations
This article presents original definitions and distinctions written for this piece rather than quotations reproduced from the dictionaries and glossaries cited below; where a source is cited, the citation supports the underlying fact or framing rather than a verbatim quotation. Usage of these terms varies across the industry, and some practitioners use them more interchangeably than the distinctions drawn here. The article does not reach a legal conclusion about the permissibility of scraping in any specific case, since that depends on the source, the data, and the jurisdiction involved.
Last reviewed: 31 July 2026.
References
- Cambridge Dictionary, "data scraping".
- National Institute of Standards and Technology, Computer Security Resource Center Glossary, "Machine-Readable", citing 44 U.S.C. Section 3502.
- Bray, T. (ed.), RFC 8259: The JavaScript Object Notation (JSON) Data Interchange Format, IETF.
- Open Data Handbook, "Scraping".
- Mozilla Developer Network, "Document Object Model (DOM)".
- Glez-Peña, D., Lourenço, A., López-Fernández, H., Reboiro-Jato, M. and Fdez-Riverola, F., "Web scraping technologies in an API world", Briefings in Bioinformatics, 15(5), 788-797.
- OpenStax, "Web Scraping and Social Media Data Collection", Principles of Data Science.
- Ferrara, E., De Meo, P., Fiumara, G. and Baumgartner, R., "Web Data Extraction, Applications and Techniques: A Survey", arXiv:1207.0246.
Disclosure: Mastros develops browser-based data-export software. This article is intended as a general technical reference and was prepared using the independent sources cited above.