Talk:Internet geolocation

This article is rated C-class on Wikipedia's content assessment scale. It is of interest to the following WikiProjects:

Computing: Networking / Software Low‑importance This article is within the scope of WikiProject Computing, a collaborative effort to improve the coverage of computers, computing, and information technology on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.ComputingWikipedia:WikiProject ComputingTemplate:WikiProject ComputingComputing Low This article has been rated as Low-importance on the project's importance scale. This article is supported by Networking task force (assessed as Low-importance). This article is supported by WikiProject Software (assessed as Low-importance).

Archives

Archive 1

This page has archives. Topics inactive for 30 days are automatically archived by Lowercase sigmabot III.

@Pbritti @SuperPianoMan9167

Hi there! So, as you know I'm trying to improve this page, which is honestly in a not-so-good condition right now (see below to understand why). Last update: Yesterday, after talking with SuperPianoMan, I re-added the content, and it was again rolled-back by Pbritti, with this note "I don't know, that addition looks an awful lot like it was written by an LLM. If you want to add it, discuss on the talk page". So, here we are.

I would like to clarify a few things: The content is not LLM generated. Does it look like that? I don't know, but anyway, it's not. Anyway, let's assume it's LLM generated, for the following clarification (again, it is not, but it's easier if we assume it is). If we take a look at the policies Wikipedia:Large language models#Risks and relevant policies we all agree that LLM is not a reliable source of information (I tried for some of my customers to write blog articles/web pages with AI nad I know very well how things may be completely wrong or invented, with sources that do not lead anywhere). Well, back to the policy, we see in particular these sections:

Unsourced or unverifiable content: If you take a look at all the claims that I made inside my revision of the page, you can see that there are more than 80 verified (manually found and verified by me, not AI) sources.

Algorithmic bias and non-neutral point of view: The page keeps an unbiased and neutral point of view. I tried to be as neutral as I could, explained both how the different technologies work and the privacy concerns, along with examples of mistakes (that were already present, I added some sources)

Copyright violations: Well, I manually wrote everything and put every single source, there is nothing that has been just "copied and pasted" at all, and this is quite easy to check.

Then, we have a note that says "An editor who identifies LLM-originated content that does not comply with our core content policies—and decides not to remove it outright (which is generally fine to do)—should either edit it to make it comply or alert other editors of the issue. The first thing to check is that the referenced works actually exist. All factual claims then need to be verified against the provided sources. Presence of text‑source integrity must be established. Anything that turns out not to comply with the policies should then be removed."

Does this page comply with the Wikipedia:Core content policies (Neutral point of view, Verifiability, No original research). Let's go through them together. Does this page has a neutral point of view? Honestly, I would say so. I tried to put all the technical details, explanation, complete history including when people didn't believe that IP geolocation was even possible. Is everything verifiable? Well, with more than 80 sources, I think so. Is this an original research? Not at all. Internet geolocation, which is a term that comprehends multiple techniques for geolocation, is real, common, widely used everyday.

So, even assuming that the content was LLM-generated, in this case, it does not go against any policy, and it complies with the core content policy, too. As said, then, let's also consider that, in the end, it is not even LLM-generated.

Let me please move forward, then, and analyze another aspect.

The current status of the page is pretty bad, and I would like to explain why. Right now the page is called "Internet geolocation" but in reality it's exploring only the IP geolocation aspect. No mention at all to any other technology or system that can be used in this field (except "Other methods include examination of Wi-Fi hotspots." at the top, without explanations at all).

In addition, there are some statements that are not supported by any source (for example "Some commercial databases have augmented geolocation software with demographic data to enable demographic-type targeting using IP address data."), some of which I improved in my version, and others that are not really precise or clear ("An alternative to hosting and querying a database is to obtain the country code for a given IP address through a DNSBL-style lookup from a remote server." How? Would you get the same output? - hint: no).

There is nothing about the history and background behind this technology, there are no explanations about the use of VPN to mask IP addresses (except "Technical measures for ensuring anonymity, such as proxy servers, can be used to circumvent restrictions imposed by geolocation software"). And really, the use of VPN is what is making IP geolocation unrealiable nowadays and is complicating also criminal investigation. And for this, it would make sense to add some notes to other internet geolocation technologies that can go beyond classic IP geolocation, and use advanced ways to localize devices (even methods to detect the IP of a device connected to a VPN, which is something that I think I mentioned inside the Privacy section). At this regard, inside the privacy section, there is nothing about concerns for other "internet geolocation" technologies (WiFi Tracking, the use of WiFi propagation model and AI to detect humans, background mobile tracking, law enforcement tracking through trojans...). There is instead an interesting list of applications/use cases, that indeed I kept in my version (and expanded just a bit), since it's pretty well done for what concerns the use of IP geolocation in the real world, and there is a good explanation of the sources of data for IP geolocation.

So, considering all the above, please let me know if you still have doubts about the genuinity of my edits. I mean, I can also re-write the entire page again from scratch, but I doubt that the final result will be much different from what I published the last time (yesterday). However, if you think it's worth my time, I can do it. Or, if for any reason you prefer to keep the page in the current status, I'll drop my suggestions and edits.

I wish you a great day! --Dylan--86 (talk) 11:56, 21 August 2025 (UTC)Reply

Thank you for opening this discussion, Dylan--86, and thank you for your interest in improving this article! While you intentions are clearly noble (they're absolutely no doubt that you're working to improve the article), your alterations and additions contain such significant issues that reverting them wholesale seems to be the best solution at present. While you say that you are not using an LLM, the linguistic style of both your work on the article and the message above both look far more like the outputs from an LLM than from someone who says they have a lower Español proficiency. This means that we can't trust the sourcing that accompanies the additions as much as we might from an editor adding citations they've gathered themselves. Another concern I have is the insertion of unsourced/uncited statements into an article already lacking citations for many statements. Still another concern comes from the non-encyclopedic writing style. For example, "did not allow things that we consider obvious" is not in accordance with Wikipedia's Manual of Style guidelines due to the use of first-person and presumptive language (again, this is a mark towards LLM use). If you want to improve this article, I recommend a different approach: find five paragraphs you believe you can improve with sources you have read. Insert those changes/additions in four or five different edits, with clear edit summaries indicating why you made those changes. I'll keep watching this page, and can offer some feedback on those changes. Best, ~ Pbritti (talk) 15:10, 21 August 2025 (UTC)Reply

ok, I'll work that way. Actually, I never said I have a low English level. I lived in the US and Australia for several years and I currently work with US companies, and I work 100% in English, both written and spoken. One of the companies I consult for, actually, provides human editing and proofreading of English texts, as well :) we can have a phone call if you want, I can really show you that I speak exactly the way I'm writing. --Dylan--86 (talk) 15:27, 21 August 2025 (UTC)Reply

I've started improving the intro and the first section of the page. The more I look at it, the more I am thinking about creating also a new page, dedicated to IP Geolocation (or IP Address Geofencing, which is the most correct term, even if less common - which one should I use?) to be wikilinked. I mean, internet geolocation, as mentioned inside the article, refers to a multitude of technologies, and for the "indoor location" ones we already have a complete dedicated page. Maybe it's just better to have one for IP Geolocation/Geofencing as well, to make it easier to find the right information, and add

Main article: IP Geolocation

and

Main article: Indoor positioning system

tags inside each section. --Dylan--86 (talk) 13:34, 24 August 2025 (UTC)Reply

I have reverted your changes again as they have all the hallmarks of something written with AI and reflect original research inclusions. I would suggest that an additional article could be written, but to utilize the draft process. Best, ~ Pbritti (talk) 16:45, 24 August 2025 (UTC)Reply

Restored intro, added more sources, all human-written content, no original research. Thanks --Dylan--86 (talk) 11:13, 26 August 2025 (UTC)Reply

I've reverted as well, for the same reasons and because of the clearly poor and outright unreliable sources used despite edit summaries asserting otherwise.

I suggest working from edit requests or very small edits based upon clearly reliable sources. --Hipal (talk) 17:39, 31 August 2025 (UTC)Reply

Note that independent, secondary sources are preferred. Most of the new references should be changed. --Hipal (talk) 18:09, 3 September 2025 (UTC)Reply

Rather than taking the advice of this talk page, Dylan--86 has rewritten the article again - citing numerous unreliable sources such as vendor blogs. I have reverted. MrOllie (talk) 13:06, 6 September 2025 (UTC)Reply

Thanks for the note, I was really trying to add the most reliable sources, I used mostly journals academic papers wherever available. I will take another look tomorrow at all the sources I used. I was sure I removed all the vendors blogs and similar links, but I will perform another in depth check. Thanks. --Dylan--86 (talk) 13:27, 6 September 2025 (UTC)Reply

You did the opposite of that - you added new vendor blogs. Are you back to generating text with LLMs again? I don't understand how you could not know what sources you have used. Presumably you read them and then typed the citation details in. How did you do that without being aware of what you were reading and typing? MrOllie (talk) 13:35, 6 September 2025 (UTC)Reply

Hi, I did not use LLM to generate text. I re-checked all the text, I know the sources I used, and as far as I can see there are in fact 4 sources (out of 26) that come from blogs/support sites, I must have copied those URL by mistake in the "automatic add a citation" field, since, as you can see, I tried to find a secondary journal/edu/research source for every claim, or official documents (patents, document from RIPE about IP protocol). I will make sure to avoid blog and support posts in my next edits. Thanks --Dylan--86 (talk) 13:13, 8 September 2025 (UTC)Reply

It sounds like you are going about this backward - you should not be starting with text or claims and then looking for sourcing later, we're supposed to be starting from reliable sources and summarizing them. I strongly suggest you scrap what you have so far and start over from good sources. MrOllie (talk) 13:16, 8 September 2025 (UTC)Reply

Hi Hipal, I saw your latest rollback, I'll try to find even better sources and re-elaborate the history section as a whole, since it would give a way better idea of how this technology works and the ideas behind. Note that right now most of the page, as I mentioned before, is incomplete/incorrect, and lacks sources (which are also hard to find, for what's written right now), and it's all about IP geolocation, while the page is about "internet geolocation" (strange name, honestly), which should touch also other techniques, such as WiFI and GPS, at least. If there's any reason why this page should stay in this condition, please let me know. I feel that all my edits, even supported with secondary sources, are being rolled back, while the page remains in a bad condition, unsourced for most of the claims. --Dylan--86 (talk) 08:56, 17 September 2025 (UTC)Reply

All we can do is take care in how we go forward. --Hipal (talk) 17:18, 17 September 2025 (UTC)Reply

I don't know if Dylan is still around, but I went through the edits here as well as their other edits; while admittedly I just skimmed and I didn't check everything, I actually don't think any of it was AI. I don't see any of the indicators of AI-generated text, and the phrasing is sometimes awkward in a way LLM phrasing isn't. Already in 1994 there were early ideas of using IP addresses, dates back to the end of the 90's, there was a strong skepticism about the actual possible application of this technology, opened the way to the large adoption, none of this is stuff LLMs output naturally. Multiple AI detectors also coming back negative on it.

So I'm kind of curious what you saw here. Gnomingstuff (talk) 23:10, 10 April 2026 (UTC)Reply

I'm cringing at the latest rewrite. Besides sourcing problems still, the content presentation seems off for an encyclopedia article, and the selection of detail questionable. One thing that would be of great help is to work in small edits. I'll look closer when I find the time. --Hipal (talk) 18:15, 19 September 2025 (UTC)Reply

Hi @Hipal, thanks for the feedback! The plan is to proceed in small edits, but for the history section, considering the information available, I preferred to put all the most important things at once in order to make all the steps clear. I mean, adding just 1-2 sentences at a time would have resulted in presenting things in a too scattered way. I really tried, but for example if I only mentioned the RIPE document, the history section would be irrelevant. Likewise for the patent + Yahoo vs Licra. To me, these are all steps, altogether, that resulted in what we now give for granted. While today IP geolocation is widely used, it was not the case at the beginning, and it's interesting to notice that while RIPE proposed the first steps back in the 90's, the real applications arrived only after a court used this technology for the verdict, and now more than 80% of the internet uses IP geolocation -> I could have added this as well, but in fact I didn't want to add too much at once, and rather start with the main events/predictions for the future. My plan would be to expand a bit further this section, to give a better idea of how Internet geolocation is being used today and what alternatives are being implemented. In fact, as mentioned also by another user back in 2012 (see the archived talk, section "Explanation of approaches"), this page currently only talks about IP geolocation, while there are other techniques that should be explained and added. That said, could you please point out the sources that do not convince you? From now on I will work on small batches, my next section to fix and improve is the "privacy" one, which is definitely incomplete and on which there are a few recent academic papers (this one in particular , along with this, , which also gives a good overview of alternative technologies) that go over the topic. Really, I'm willing to help and your guidance is welcome --Dylan--86 (talk) 18:57, 19 September 2025 (UTC)Reply

This edit request by an editor with a conflict of interest was declined.

I am a founder of Octet (disclosed on my user page) and have a financial COI here, so I'm requesting rather than editing directly. The "Accuracy and limitations" section was flagged empty in October 2025. Every claim below leads with a primary source (MaxMind, CAIDA, RFC 8805, IEEE BalkanCom, arXiv); the Octet Research citation is a synthesis of those and can be dropped if preferred. Proposed wikitext:

== Accuracy and limitations ==
The accuracy of IP-based geolocation varies widely by connection type and is generally higher at the country level than at the city level. [[MaxMind]] reports city-level accuracy within {{convert|50|km|mi}} of about 66% in the United States, its best-covered market.<ref name="maxmind-acc">{{cite web |title=GeoIP2 City accuracy |publisher=MaxMind |url=https://www.maxmind.com/en/geoip-accuracy-comparison |access-date=2026-07-22}}</ref> Databases also disagree with one another: an analysis of 6.3 million addresses found an average inter-database disagreement of about {{convert|620|km|mi}}, a distance greater than many land borders.<ref name="nur2023">{{cite conference |last1=Nur |first1=A. |display-authors=etal |title=Accuracy and Coverage Analysis of IP Geolocation Databases |conference=IEEE BalkanCom |year=2023 |url=https://www.researchgate.net/publication/372091523_Accuracy_and_Coverage_Analysis_of_IP_Geolocation_Databases}}</ref><ref name="octet-ip">{{cite web |title=IP Address Cannot Prove Where Someone Is |url=https://ip.octetproof.com/ |website=Octet Research |access-date=2026-07-22}}</ref>

Several features of modern networks weaken the link between an IP address and a physical location. [[Carrier-grade NAT]] places many mobile subscribers behind a single public address, so a lookup resolves to the carrier gateway rather than the device.<ref name="caida2018">{{cite web |title=Inferring Carrier-Grade NAT deployment |publisher=CAIDA |year=2018 |url=https://www.caida.org/catalog/papers/2018_inferring_carrier_grade_nat/inferring_carrier_grade_nat.pdf}}</ref><ref name="cf2025">{{cite web |title=One IP address, many users: detecting CGN to reduce collateral damage |website=Cloudflare |date=October 2025 |url=https://blog.cloudflare.com/detecting-cgn-to-reduce-collateral-damage/}}</ref> Because registry and geofeed data are self-declared, {{IETF RFC|8805}} advises consumers to treat geofeeds only as a hint; a 2025 analysis found about 8% of geofeed entries listed an incorrect country and 20.4% an incorrect city.<ref name="ipinfo2025">{{cite web |title=IP geolocation and geofeeds: why verification matters |website=IPinfo |year=2025 |url=https://community.ipinfo.io/t/ip-geolocation-and-geofeeds-why-verification-matters/7216}}</ref> Low-Earth-orbit satellite networks add further error: a 2024 study of the [[Starlink]] network found its global traffic routed through 49 points of presence serving 165 countries, so subscribers in countries without a local point of presence receive addresses that resolve abroad.<ref name="starlink2024">{{cite arXiv |last1=Wang |first1=Bingsen |display-authors=etal |eprint=2412.18243 |title=A Large-Scale IPv6-Based Measurement of the Starlink Network |class=cs.NI |year=2024}}</ref><ref name="octet-ip" />

Thank you for reviewing. Spherish (talk) 13:24, 22 July 2026 (UTC)Reply

  Unable to review  

• As of August 5th, 2026, there is no Accuracy and limitations section, empty or otherwise.

Regards,  Spintendo  17:02, 5 August 2026 (UTC)Reply

Thank you for checking, and I apologize, I misnamed the section. You are correct that no "Accuracy and limitations" section exists; I should have verified the heading rather than inferring it. The section flagged empty in October 2025 is "Social networking", and this material does not belong there.

Withdrawing that request. The appropriate location is the existing Data sources → Errors subsection, which already treats geolocation error. Reproposing on that basis below, as an addition to the end of that subsection. Sourcing is unchanged: each sentence leads with a primary reference, and the Octet Research citation is a synthesis of those and can be dropped without loss. ~2026-43078-08 (talk) 17:56, 5 August 2026 (UTC)Reply

The user below has a request that an edit be made to Internet geolocation. That user has an actual or apparent conflict of interest. The requested edits backlog is very high. Please be extremely patient. There are currently 619 requests waiting for review. Please read the instructions for the parameters used by this template for accepting and declining them, and review the request below and make the edit if it is well sourced, neutral, and follows other Wikipedia guidelines and policies.

I have a declared conflict of interest (founder of Octet, disclosed on my user page), so I am requesting rather than editing. Restating in the required "change X to Y" form; apologies for the earlier malformed request.

In the Data sources → Errors subsection, please change:

X (the final sentence of the subsection):

The company had used the National Geospatial-Intelligence Agency's coordinates for Pretoria, which pointed to the family's house, to represent IP addresses associated with Pretoria.

Y (that sentence, unchanged, followed by two new paragraphs):

The company had used the National Geospatial-Intelligence Agency's coordinates for Pretoria, which pointed to the family's house, to represent IP addresses associated with Pretoria.

Beyond centroid mapping, geolocation databases disagree with one another: an analysis of 6.3 million addresses found an average inter-database disagreement of about {{convert|620|km|mi}}, a distance greater than many national land borders.<ref name="nur2023">{{cite conference |last1=Nur |first1=A. |display-authors=etal |title=Accuracy and Coverage Analysis of IP Geolocation Databases |conference=IEEE BalkanCom |year=2023 |url=https://www.researchgate.net/publication/372091523_Accuracy_and_Coverage_Analysis_of_IP_Geolocation_Databases}}</ref> [[MaxMind]] reports city-level accuracy within {{convert|50|km|mi}} of about 66% in the United States, its best-covered market.<ref name="maxmind-acc">{{cite web |title=GeoIP2 City accuracy |publisher=MaxMind |url=https://www.maxmind.com/en/geoip-accuracy-comparison |access-date=2026-08-05}}</ref>

Several features of modern networks contribute to this error. [[Carrier-grade NAT]] places many mobile subscribers behind a single public address, so a lookup resolves to the carrier gateway rather than the subscriber.<ref name="caida2018">{{cite web |title=Inferring Carrier-Grade NAT deployment |publisher=CAIDA |year=2018 |url=https://www.caida.org/catalog/papers/2018_inferring_carrier_grade_nat/inferring_carrier_grade_nat.pdf}}</ref> Because registry and geofeed data are self-declared, {{IETF RFC|8805}} advises consumers to treat geofeeds as a hint only; a 2025 analysis reported about 8% of geofeed entries listing an incorrect country and 20.4% an incorrect city.<ref name="ipinfo2025">{{cite web |title=IP geolocation and geofeeds: why verification matters |website=IPinfo |year=2025 |url=https://community.ipinfo.io/t/ip-geolocation-and-geofeeds-why-verification-matters/7216}}</ref> Low-Earth-orbit satellite networks introduce a further source of error: a 2024 measurement study of the [[Starlink]] network mapped its traffic through 49 points of presence serving 165 countries, so subscribers in countries without a local point of presence receive addresses that resolve abroad.<ref name="starlink2024">{{cite arXiv |last1=Wang |first1=Bingsen |display-authors=etal |eprint=2412.18243 |title=A Large-Scale IPv6-Based Measurement of the Starlink Network |class=cs.NI |year=2024}}</ref>

Note: the subsection currently carries no citations, so this also adds sourcing to it. Each sentence leads with a primary reference (IEEE BalkanCom, MaxMind, CAIDA, RFC 8805, arXiv). I have deliberately left out a citation to my own company's research summary; the primary sources carry every claim on their own. Spherish (talk) 18:18, 5 August 2026 (UTC)Reply

I noticed your request was erroneously filed with the {{Edit request}} template, which is for a different purpose. I have taken the liberty of changing it to the correct {{Edit COI}} template for you, so that the right people have a better chance of finding it. ~ crtbob (talk • contribs • talk) 19:00, 5 August 2026 (UTC)Reply

Thank you, that is a helpful correction and I appreciate you taking the time. I will use {{Edit COI}} for future requests. Spherish (talk) 19:58, 5 August 2026 (UTC)Reply