An IP ban is a rule on a server or firewall that refuses every request coming from one IP address, or from a range of addresses. The site usually answers 403, or does not answer at all, on every URL, and waiting a minute changes nothing. Three other refusals look similar and are resolved differently: a rate limit, an account block and a geo block.
People say "I got IP banned" for all four. The response usually tells you which one it is, and that decides what to do next. This post shows how to read it, with a small classifier we ran on 20 November 2026, and what legitimately resolves each case when you collect public data.
What an IP ban is
Every request reaches a site from a public IP address. A ban is a list entry that says: refuse this address. It can live in the web server, in a firewall, or at the content delivery network in front of the site. Three details matter in practice.
- It can cover one address or a range. Operators often ban a whole block when the requests they object to come from many neighbouring addresses.
- It can be temporary or permanent. Automatic bans placed by tools that watch for repeated failed logins or bursts of requests usually expire after minutes or hours. Bans placed by a person stay until a person lifts them.
- It applies to everyone behind the address. A home router gives one public address to every device in the house, and mobile carriers put many subscribers behind one address. A ban on that address reaches all of them, which is also why sites use bans with care.
Sites ban addresses that send far more requests than a person would, that ignore robots.txt, that flood login or search forms, or that were reported for abuse. A ban is the site saying no to the address, not to one request.
Four refusals that look alike
| Refusal | Tied to | Usual status | What the response says | Does waiting help? |
|---|---|---|---|---|
| Rate limit | Requests per period, per address, session or key | 429 | Often a Retry-After header, or "too many requests" | Yes, after the stated time |
| IP ban | The address or its range | 403, or no answer | "Access denied", sometimes "your IP has been blocked"; the same on every URL | Only if the ban is temporary |
| Account block | The login, token or API key | 401 or 403 | Names the account: suspended, disabled, invalid token. Logged-out pages still load | No: it goes through the platform's support |
| Geo block | The country of the address | 451, or 403 | "Not available in your country or region", or a redirect to another country's site | No: it is a rule about location |
The four differ in what they are tied to. A rate limit is about pace and ends on its own. An IP ban is about the address. An account block follows the login to any address. A geo block applies to every visitor from that country, however they behave.
Read the response before anything else
The status code. 429 is defined for one purpose, too many requests, so it is the clearest signal of the four; 429 Too Many Requests covers it in detail. 451 means "unavailable for legal reasons" and is the standard way to say a page is withheld in a country. 401 asks for valid credentials. 403 is the general "understood, refused" and needs the next two readings; our post on 403 Forbidden lists what else can cause it.
The headers. A Retry-After header means the refusal has an end, which a ban does not announce; how to read it is in the post on the Retry-After header. Many services also publish their limits on every answer, before you reach them. On 20 November 2026 a call to GitHub's public API without a token came back with these:
$ curl -s -i https://api.github.com/rate_limit
HTTP/1.1 200 OK
X-RateLimit-Limit: 60
X-RateLimit-Remaining: 60
X-RateLimit-Resource: core
Sixty requests an hour for that address, all sixty still available. A client that reads X-RateLimit-Remaining knows where the limit is and never has to meet it. The mechanics behind such limits are in our explainer on rate limiting.
The page. Block pages from large networks say which rule fired. Cloudflare error 1015 reads "You are being rate limited": a rate limit set by the site owner. Cloudflare error 1020 reads "Access denied": a firewall rule, which can be an address, a country or something else about the request. Akamai's Access Denied page carries a reference number. Keep that number or the Ray ID: it is what the site's administrator needs to find the rule that matched. Not every refusal arrives with an error status; the four reply shapes we measured on 13 sites shows what the others look like.
Then three quick observations settle most cases:
- Does every URL of the site answer the same way, including the home page and robots.txt? A ban covers the whole host; a rate limit or a permission problem often covers one path.
- Does the answer mention an account, a token or a login? Then the address is not the subject.
- Does it mention a country or region? Then it is a location rule: see what geo-blocking is, and our measurements of the same URL from different countries.
A small classifier, tested
The same reading as code. It looks only at the response, in the order above: status, headers, text.
GEO_WORDS = ("your country", "your region", "your location", "not available in")
ACCOUNT_WORDS = ("account", "suspended", "token", "sign in", "log in")
def classify(resp):
"""Name the refusal from the response alone: status, headers, then the text."""
body = resp.text[:2000].lower()
if resp.status_code < 400:
return "ok"
if resp.status_code == 429 or "Retry-After" in resp.headers:
return "rate limit: wait %s s, then slow down" % resp.headers.get("Retry-After", "a while")
if resp.status_code == 451 or any(w in body for w in GEO_WORDS):
return "geo block: decided by the country of the IP"
if resp.status_code == 401 or any(w in body for w in ACCOUNT_WORDS):
return "account block: tied to the login, not the address"
if resp.status_code == 403:
return "possible IP ban: stop, check robots.txt and the terms, ask the site"
return "other error %d" % resp.status_code
We ran it on 20 November 2026 (Python 3.12, requests) against a local test server that answers each case, so no real site was asked to refuse anything:
/rate 429 rate limit: wait 30 s, then slow down
/ban 403 possible IP ban: stop, check robots.txt and the terms, ask the site
/account 403 account block: tied to the login, not the address
/login 401 account block: tied to the login, not the address
/geo 451 geo block: decided by the country of the IP
/geo403 403 geo block: decided by the country of the IP
/ok 200 ok
Seven responses, seven correct names. Two of the 403s were not bans at all: one named a suspended account and one named the visitor's region. That is the point of reading the text before deciding. The word "possible" on the ban line is deliberate: a bare 403 can also be a decision about the request itself, and what confirms a ban is that every URL answers the same way and the answer does not change with time.
What resolves each one
Rate limit. Wait for the time in Retry-After, then send fewer requests: one at a time per host, with a pause between them, and no refetching of pages you already have. It clears by itself, and it comes back if the pace does.
IP ban. Stop sending requests to that site. Then check two documents: its robots.txt, which you can test path by path with our robots.txt tester, and its terms of use. What they say decides the next step.
- If the site does not allow automated access to those pages, the ban is the site enforcing its rules, and the answer is to stop. Look for an official API, a data licence or a public dataset of the same information.
- If your collection is within the rules, or you were only browsing, the ban may be a mistake or may have been earned by someone else on a shared address. Write to the site. Give your public IP address, the date and time, the URL, and the reference or Ray ID from the block page. Say what you collect, how often, and how to reach you. Administrators can lift a ban or add an address to an allow list, and can point you to the feed or API they prefer you to use.
- If the ban is automatic and temporary, it expires. Come back at a lower pace, or it will be placed again for longer.
Account block. Only the platform can lift it, through its support or appeal form. It is tied to the account, not to the address.
Geo block. The content is not offered where the request comes from. For services restricted by licence or law, that is the end of it. Public pages that simply differ by market, such as prices or search results, are a different matter: that is a local version, not a refusal.
Rotating or one address: a setting, not a remedy
When you collect public pages through a proxy network you choose how addresses are assigned. With rotating proxies each request goes out on the rotating line; with a named session, the session name and its length are set in the username. It is a choice about the shape of the job:
- Rotating suits independent requests, where each page is fetched once and no state links one to the next: a list of product pages, a set of search queries.
- A named session suits steps that belong together, such as a search followed by its result pages, or anything that depends on a cookie set by the first request.
We checked both on the QuanticData Residential Basic network on 20 November 2026, country set to the United States: 10 requests on the rotating line came out of 10 different addresses, and 10 requests on one named session came out of the same address, 10 of 10. The session name and its length are part of the proxy username, so the choice is made per job. Details are on the residential proxies page.
Neither setting changes what a site allows. The site's limits apply to the whole of your traffic, so the pace you keep towards one host is the total of all your requests, whatever addresses they leave from. A 403 or a 429 means slow down or stop on that site, not move.
How not to earn a ban
Most bans on data collection follow the same few habits, and the opposite habits prevent them:
- Read robots.txt before the first page, and skip what it disallows.
- Keep to about one request per second per host, and slower on small sites.
- Send a User-Agent that says who you are, with a contact address, so an administrator can write to you before blocking.
- Cache, and use conditional requests, so unchanged pages are not downloaded twice.
- Prefer the site's API or feed when it has one.
- Treat 429 and 403 as instructions. Retries are for timeouts and for 502, 503 and 504.
A complete polite crawler built on these rules, with the code and a real run, is in a separate post. If you would rather not run the fetching yourself, the QuanticData web scraping API takes a URL and returns the page as Markdown or HTML.


