On this page you will find the answers to the following questions:

• What does "tweeps" mean?
• How does Tweeps.net work? (And how doesn’t it work?)
• Why do I get an "Interface constraint encountered" error message?
• How is bot analysis done?
• Why aren’t all followers or people followed analysed?
• What aspects of tweets are analysed?
• What do bot scores mean?
• How should bot scores be interpreted?
• What do troll bot, spam bot, other bot, troller, spammer and unclear mean?
• What does the tweets / hours chart mean?
• What is the veil of followers?
• Who created Tweeps.net?

What does "tweeps" mean?

In the singular form, tweep means X (formerly Twitter) user in English. (Source: Cambridge Dictionary)

Tweeps is the plural form and refers to either X users or followers of a particular user on X. (Source: Lexico)

How does Tweeps.net work? (And how doesn’t it work?)

Tweeps.net seeks to identify bots based on two main principles:

• Based on machine-like operation
• Based on activities that are atypical for human users

Tweeps.net does NOT identify bots based on the items below because they are not suitable for it:

• Based on following popular users (because X automatically recommends them to new users)
• Based on inactivity (because in that case there is no evidence of automatic activity)
• Based on the absence of a profile picture (although this information is displayed for others identified as bots)

In short, bots are identified based on the performance of their X IDs, and bot scores are not affected by a profile’s external factors.

Why do I get an "Interface constraint encountered" error message?

If you receive an "Interface Restriction Encountered" error message, it is because the X interface (API) is currently blocking retrieval of data from users. This error should go away by itself after a few minutes.

By logging in with your own X account, you can do more searches more often.

How is bot analysis done?

This is how Tweeps.net analyses bots:

1. Searches for an account’s followers or people followed.
2. Retrieves profile information of followers or people followed.
3. The aim is to predict suspicious accounts based on profile data. This uses previous correlation analysis to see which types of accounts are most likely to receive bot scores. In addition, those previously identified as bots (a bot score higher than 2) are taken into account.
4. Retrieves tweets of suspicious accounts (200 most recent).
5. Analyses tweets of suspicious accounts. Bot scores are calculated at this point. Bot scores are therefore given only for tweets, i.e., the active activity of an account.

At the same time, the following are taken into account:

• Banning of the account from the X search function
• Suspension or blocking of an account by X. These are accounts that are often deleted later.
• Blocking between accounts
• Protected accounts cannot be analysed if the person performing the analysis has not been accepted as a follower by that user

Why aren’t all followers or people followed analysed?

Tweeps.net operates with the X software interface (API). The interface has a search volume limit per 15 minutes. For this reason, not all followers or peopled followed can be retrieved; there may be tens of thousands of them. It could be possible, but it would take a very long time.

Secondly, tweets are analysed only if a suspicious user has at least 200 tweets. The reason for this is to get a sample large enough with respect to the reliability of the results.

Third, the number of accounts to be analysed has been artificially limited so that the analysis does not take too long. The analysis has been designed to focus on those that are likely to receive the highest bot scores. It wouldn’t be very interesting to go through hundreds of potential bots with scores of 1-2 points.

What aspects of tweets are analysed?

Tweeps.net calculates bot scores based on about 20 indicators that are based on the following, among others:

• Tweeting hours, i.e., how many different hours of the day are tweets of the analysed account produced.
• Patterns identified in tweeting hours
• Activeness, number of tweets per hour
• On how many days the tweets of an analysed account occur.
• Tweet lengths
• Types of tweets: regular, replies, quotes and retweets
• Links, photos and videos in tweets
• #hashtags in Tweets
• @mentions of other users in tweets (without being a reply tweet)
• Similarity of the contents of tweets
• X applications used to publish tweets

What do bot scores mean?

Bot scores are divided into the following types:

• Machine-like

For humans, it is unusual to send tweets regularity or focus on certain hours.

• Tireless

The user has tweeted around the clock at least once in the previous week.

• Like clock-work

Tweets aren’t spread across the day, but only take place during a few hours.

• Mechanical approach

Tweeting times appear to follow at least a partially repetitive pattern.

• Activeness

Tweeting activity is unusually high.

• @-spam

The tweets contain unusually frequent mentions of other users without being replies. This may indicate harassment.

• Reply spam

Tweets are unusually frequently replies to other users' tweets.

• Media spam

Tweets contain an unusual number of photos and / or videos.

• Link spam

Tweets contain an unusual number of links.

• Hashtag spam

Tweets contain an unusual number of hashtags.

• RT spam

Tweets are retweets to an unusually large extent.

• Quote spam

Tweets contain an unusual amount of quotes from other tweets. This may indicate online targeting and shaming.

• Spam campaign

The tweets are similar to that of several other users. This suggests that the user is part of a spam campaign.

• Adult content spam

The user often shares links to content identified as sensitive by X. Content Warning.

• Repeated length

The length of tweets is often about the same.

• Duplicate content

Tweets often repeat the same content.

• Repetitive RT content

Retweets often repeat the same content.

• Automatic app

The tweets were published using an automatically functioning X application, which is not considered suspicious.

• Suspicious app

The tweets were published using an automatically functioning X application, which is considered suspicious.

In addition, the following aspects are displayed but do not affect bot scores:

• Default profile

The account has a default profile picture and has not changed the look and feel of their profile page.

• Adult content

X has identified sensitive content in links shared by the account. Content Warning.

• Banned from search

The account has been banned from X search. Therefore, the account's Tweets will not appear in the search function.

How should bot scores be interpreted?

Bot scores are not direct proof of whether an account is really a bot or not. This is because sometimes people act in a mechanical or otherwise bot-like way. When an account shows a high bot score, there should be no ambiguity, nevertheless we use the term "potential bots". Below is an interpretation guide.

• 1-2 points: It is not clear whether the user is a human or a bot. This may be due to an unusually high level of activity or large number of retweets.
• 3-4 points: The user is either a bot or a person whose tweeting activity is unusual. The explanation may be spamming and tweets being posted during only a few hours of the day.
• 5-10 points: The account is a typical possible bot. The score is explained by a number of different factors such as activity, tweeting pattern, spam content, or repetitive content.
• More than 10 points: This is very clearly a probable bot.

In addition to bot scores, you should also look at the tweets / hours chart.

What do troll bot, spam bot, other bot, troller, spammer and unclear mean?

For each X account that receives bot scores, an automatic interpretation by the algorithm based on the types of points is displayed.

The automatic interpretation of bot scores can be one of the following: troll bot, spam bot, automatic app, bot app, other bot, troll, or spammer. If there are too few bot scores to make an interpretation, the text "unclear" is displayed.

The types of bot scores affect the interpretation is shown in the following table:

Types of bot scores		Human	Social bot
Machine-like Tireless Like clock-work Mechanical approach Activeness Repeated length Duplicate content Repetitive RT content		-	Other bot
+	Automatic app Suspicious app		Automatic app/ Bot app
+	Media spam Link spam Hashtag spam RT spam Spam campaign Adult content spam	Spammer	Spam bot
+	@-spam Quote spam Reply spam	Troll	Troll bot

 

What does the tweets / hours chart mean?

In addition to bot scores, the tweets / hours chart is displayed for accounts and potential bots. This is called a polar chart. Here is an example:

The chart shows how the tweets posted by an account and included in the analysis are distributed over the hours of the day. Sector 0 means tweets published between 0-1 a.m., sector 2 means tweets published between 2 and 2 p.m., and so on. So the picture shows what time the user has tweeted.

For example, in the image above, the account has tweeted all of its 200 tweets included in the analysis between 9pm and 10pm. Usually, tweets by humans are spread out over different hours of the day.

Below the graph is shown how on many days the 200 tweets included in the analysis are posted, as well as the date of the last published tweet. This information makes it possible to deduce how active the account has been recently.

It is common for bots to have tweets focused on either a small number of hours or during particular hours. Here are some examples, some of which clearly indicate a regular pattern in the times when tweets are posted:

What is the veil of followers?

The veil of followers is presented at the end of the bot analysis. Here is an example:

The points in the graph are followers of an account. The vertical axis represents date when followers of a X account were created, and the horizontal axis represents the following order. The last followers of the account are on the chart on the right and the first on the left.

Sometimes, a lot be deduced from this graph, but here are the main points:

• Based on the analysis, the red dots are potential bots, i.e. accounts which received at least one point in the bot analysis.
• Normally, bots can be expected to be evenly distributed in the horizontal axis. If there are a lot of bots at some particular point, the account has gained more potential bot followers than usual at that point.
• Usually, accounts gain new followers who have joined X recently at a regular pace. The chart can therefore be expected to rise steadily.
• Changes in the "slope" of the chart indicate a change in the rate at which new followers accumulate.
• If the account is found to have received an unusually large number of new followers who have joined X on the same day, they will be displayed on the chart as green dots with an explanation of "suspicious due to the date of joining."
• In some years, more users have joined X than in others. Such years may differ in the veil graph as point densities in the horizontal direction.
• X has removed thousands of troll and bot accounts from the service. As a result, the veil graph may have periods with fewer points indicating followers than elsewhere. This can, of course, be evident for accounts with a very large number of followers.

Who created Tweeps.net?

The creator and administrator of the service is Tweeps Oy, based in Oulu, Finland. The bot analysis tool was developed by Harto Pönkä (@hponka). He has extensive experience in performing X analyses and learning how bots operate.