Is Web Scraping Legal in 2026? A Practical Guide to GDPR, CFAA and Terms of Service
This article is general information, not legal advice. Laws differ by country and change over time, and the facts of your project matter. Ask a qualified lawyer before you start anything that touches personal data, logins or a contract you have signed.
With that said, the question "is web scraping legal" has a more useful answer than "it depends". For most business projects, the answer turns on three things: what kind of data you collect, how you reach it, and what you do with it afterwards. This guide walks through each one and uses a real example, our free Product Hunt launches sample, to show what a low-risk dataset looks like.
The short answer
Collecting publicly visible, non-personal facts such as prices, product names, ratings and listing counts is common business practice and is generally treated as lower risk than other kinds of scraping. The risk grows when the data is personal, sits behind a login, is protected by technical access controls, or is copied wholesale and republished as your own.
Web scraping legality is not one law. It is a stack of separate questions:
- Data protection law (GDPR in Europe and the UK, similar laws elsewhere): is the data about identifiable people?
- Computer access law (the CFAA in the US, comparable statutes elsewhere): are you accessing something you are not authorised to access?
- Contract law: did you agree to terms of service that forbid automated access?
- Copyright and database rights: are you copying protected content rather than facts?
You can pass one test and fail another, so check all four.
Public data vs personal data
The most useful line to draw is between data about things and data about people.
Data about things includes product prices, stock status, tagline text, category tags, vote counts, hotel rates and business opening hours. Nobody's private life is in it.
Data about people includes reviewer names, profile photos, personal e-mails, private phone numbers, social handles of individuals and anything that lets you identify or contact a private person. Under GDPR this is personal data, and "it was publicly visible" is not a free pass. You still need a lawful basis, a clear purpose, data minimisation and a way to honour deletion requests.
This is why GDPR web scraping questions are mostly about fields, not about the act of fetching a page. A scraper that collects only product facts has very little GDPR exposure. A scraper that collects author names, photos and profile links from the same page has a lot.
A practical rule we follow: if a column would let someone identify a private individual, leave it out unless there is a clear lawful basis and a real need. Business names and business contact details on public directory listings are treated differently from private individuals, but you should still check the rules in your market.
The CFAA and "authorised access"
In the United States, the Computer Fraud and Abuse Act is the statute people usually mention. Court decisions over the last few years have leaned toward the view that accessing pages open to anyone on the public web is not "without authorisation" in the CFAA sense. The same decisions are far less friendly to scraping behind a login, using someone else's credentials, or getting around technical barriers designed to keep you out.
The practical takeaways:
- Do not scrape pages that require sign-in. If the data is only visible to logged-in users, treat it as off limits unless you have explicit permission.
- Do not use stolen or shared credentials, ever.
- Do not overload a site. Slow, polite request rates avoid harm and avoid claims that you damaged a service.
- Other countries have their own rules. The absence of a CFAA problem does not settle the question in the EU, the UK, India or Turkey.
Terms of service and contracts
Many sites forbid automated collection in their terms. Whether those terms bind you depends on whether you actually agreed to them. Clicking "I agree" when creating an account is a clear agreement. Simply visiting a public page is a weaker basis for a contract claim, though courts differ and sites can still block you technically.
The practical position: read the terms of any site you plan to collect from regularly, never create an account just to scrape, and ask whether an official API or a data licence would be better. We cover that trade-off in web scraping vs API, and the basics of what is feasible in can you scrape data from a website.
Copyright and database rights
Facts are not copyrightable in most jurisdictions, but creative text, images and sometimes the structure of a database are protected. Collecting a product price is different from copying a full article or a library of photos. Some regions, notably the EU, also have sui generis database rights that can apply to substantial extractions.
Safer habits: collect facts and short metadata rather than full creative content, keep your own derived analysis rather than republishing the source, and credit sources where appropriate.
What our Product Hunt sample shows
To make this concrete, here is what a low-risk dataset looks like. Our Product Hunt launches sample holds 112 products from the public daily leaderboards for 19 to 25 October 2026. Every column is a fact about a product, not a person: launch date, rank, product name, tagline, topics, upvotes, comments, product URL and a collection timestamp.
We deliberately left out maker names, commenter names and profile photos. They are not needed to answer the questions buyers ask, and leaving them out removes most of the personal-data risk.
Figures computed from the file:
- Upvotes: the average product has 146.4 upvotes and the median is 110. The range runs from 72 to 493.
- Comments: the average is 14.0 per product and the median is 8.
- Concentration: the top five products of each day hold 51.5% of all upvotes in the sample.
- Rank curve: median upvotes fall from 348 at rank 1 to 166 at rank 5, 103 at rank 10 and 92.5 at rank 14.
- Topics: 51.8% of products carry the Artificial Intelligence tag (58 of 112). Developer Tools appears on 42 products and Productivity on 33.
- Topic strength: products tagged Artificial Intelligence have a median of 126 upvotes, against 118 for Developer Tools and 98 for Productivity.
- Top launch: the highest single count was 493 upvotes, on 20 October.
The 25 October leaderboard was captured while the day was still running, so it has 13 rows and the lowest daily upvote total (1,434). Complete days range from 1,920 (a Saturday) to 2,825.
An excerpt of the five top products from 23 October:
| launch_date | rank | product_name | upvotes | comments |
|---|---|---|---|---|
| 2026-10-23 | 1 | Product Hunt Visual Launch Editor | 427 | 23 |
| 2026-10-23 | 2 | MOVO AI | 297 | 45 |
| 2026-10-23 | 3 | Agent Hunt | 268 | 7 |
| 2026-10-23 | 4 | Ovren Live Preview | 197 | 53 |
| 2026-10-23 | 5 | ElevenLabs Realtime | 187 | 3 |
Why is this a safe example? The pages are public with no sign-in. The fields describe products and public counts. No private individual can be identified from the file. And the data answers a real business question, namely how competitive a launch day is.
Which fields matter, and which to avoid
When you design a dataset, review every column against three questions:
- Do we need it? If a field does not serve the use case, drop it.
- Can it identify a private person? If yes, stop and get advice.
- Is it a fact or creative work? Prefer facts.
For a launch dataset, the useful fields are rank, topics, upvotes and comments. For a price-monitoring job, they are price, availability and seller. For a lead list, business name and business phone from public directories may be fine in some markets, while personal mobile numbers are not.
Common practical challenges
Legal care is only part of a scraping project. Teams also run into:
- Anti-bot measures. Rate limits, CAPTCHAs and fingerprinting. Respecting a site's limits and using managed infrastructure reduces both errors and friction.
- Pagination and date ranges. Leaderboards, listings and search results need consistent paging so you do not miss or duplicate rows.
- Layout changes. Sites change markup without notice, so parsers need regular checks.
- Data quality. Duplicates, partial days (like our 25 October capture), missing values and inconsistent formats need cleaning and a documented method.
A compliance checklist you can reuse
- Write down the purpose of the project in one sentence.
- List the fields and mark any that relate to people.
- Remove personal fields you do not need.
- Confirm no login is required.
- Read the site's terms and robots rules, and note them.
- Keep request rates gentle.
- Store the collection date and method with the data.
- Set a retention period and delete what you no longer need.
- Have a lawyer review anything involving personal data, health or financial information, or regulated sectors.
Ethical web scraping is mostly this: collect what you need, collect it politely, leave people out, and be transparent about the method.
How often should you refresh?
It depends on how fast the data changes. Launch leaderboards change daily and finish within a day, so a daily snapshot after the day closes is enough. Prices may need hourly or daily checks. Business directory data is fine on a monthly or quarterly cycle. Fewer, cleaner refreshes are better than constant ones; they also reduce load on the source.
FAQ
Is web scraping legal? Scraping public, non-personal data is common and generally lower risk, but legality depends on the data type, how you access it, the site's terms and your country's laws. Get legal advice for your specific case.
Does GDPR apply if the data is public? Yes, if it relates to identifiable people. Public visibility does not remove GDPR duties. Avoiding personal fields is the simplest way to stay outside the problem.
Can I scrape a site that forbids it in its terms? Terms can create contract risk, especially if you accepted them with an account. Review them, consider an official API or licence, and ask a lawyer when the data matters to your business.
Is it okay to scrape behind a login? We do not do it. Login-protected pages raise access-authorisation and contract issues, and we skip any target that requires sign-in.
Next steps
If you want to see what a compliant dataset looks like, download the Product Hunt sample and check each column for yourself. If you have a target of your own, request a free sample of your own target and we will tell you honestly which fields we can collect and which we would leave out. Datasets start from $49; see pricing.
Author: DataHarbor Team