Fraud Score IP: what threshold is needed for social networks and stores

Accounts are not always banned for suspicious behavior — often the reason lies in the IP address itself. Fraud Score is an assessment of the "fraudulence" of an IP from 0 to 100, which platforms use to decide whether to trust the user or send them to a shadow ban. We will analyze what threshold is safe for social networks, marketplaces, and scraping, and how to avoid falling into the risk zone.

What is Fraud Score IP

Fraud Score (sometimes referred to as Trust Score in reverse scale) is a numerical assessment generated by anti-fraud systems based on the analysis of an IP address. The higher the value, the greater the likelihood that a bot, spammer, fraudster, or user attempting to bypass platform restrictions is hiding behind that IP. The assessment is used not only by payment systems but also by social networks, marketplaces, and advertising platforms like Facebook Ads and TikTok Ads.

The practical value of the metric is that it provides an objective number where previously one had to guess: "why did the account get a shadow ban?", "why does Wildberries ban the scraper?", "why did the ad account go into review 20 minutes after launch?". Often the answer is simple — the proxy's IP had a Fraud Score above 50-60, and the system automatically classified the session as suspicious.

Services like IPQualityScore, Scamalytics, and ProxyScore provide public access to check this parameter. This is why arbitrage specialists and SMM professionals first run a batch of IPs through such checks before purchasing a proxy pool, rather than immediately uploading them to an anti-detect browser.

How the score is calculated

The calculation algorithms are closed, but based on the open data from anti-fraud providers, we can identify the main factors that increase the Fraud Score:

Type of IP — data center addresses by default receive higher scores than residential or mobile ones, as data centers are widely used for bots and scrapers.
Usage history — if spam was previously sent from this IP, likes were artificially inflated, or fake accounts were created, the score increases even after the address changes ownership.
Geolocation match with the device's time zone — discrepancies between IP geolocation and the device's system settings are a red flag for anti-fraud systems.
Connection density — if hundreds of different accounts log in simultaneously from one IP (as happens with cheap shared proxies), this sharply increases the score.
Presence in blacklists — databases like Spamhaus, SORBS, and similar lists automatically raise the Fraud Score.
ASN provider — some hosting providers and VPN services are known to anti-fraud systems as sources of mass fraud, and the entire range of their IPs receives penalties.

An important nuance: Fraud Score is not a fixed property of an IP address forever. It is recalculated dynamically and can change throughout the day, especially for residential pools with rotation.

Decoding the scale 0-100

There is no universal standard, but most anti-fraud services and platforms rely on a similar gradation:

Range	Risk Assessment	What it means in practice
0-25	Low	The IP appears as a regular home or mobile user. Passes almost all checks.
26-50	Medium	Acceptable for most tasks, but additional checks (captcha, SMS confirmation) may occur.
51-75	High	Risk of shadow banning, ad account blocking, registration denial.
76-100	Critical	The IP is almost certainly blacklisted, access is instantly blocked on most platforms.

The problem is that the "safety" threshold varies for different tasks. Registering a new account on Instagram requires a stricter filter than simply reading pages on Wildberries for price monitoring. We will break this down by segments below.

Social networks are the most sensitive platforms to Fraud Score because fake accounts and bots are their main headache in terms of moderation and advertisers. The practice of arbitrage specialists shows the following benchmarks:

Facebook Ads / account registration — a Fraud Score no higher than 10-15 is needed. Facebook aggressively bans new accounts from suspicious IPs already at the registration stage, before any advertising activity.
Instagram (farming and posting) — an acceptable threshold is up to 20-25, but for large-scale multi-accounting (30+ profiles), it is better to stay below 15; otherwise, some accounts will end up in shadow ban in the first week.
TikTok Ads — here the anti-fraud system is more strictly tied to the mobile network operator. The working threshold is up to 10, and it is important that the geolocation of the IP matches the targeting of the campaign.

For social networks and advertising accounts, mobile proxies yield the best results — they use IPs from real mobile network operators, and their Fraud Score is almost always in the range of 0-10 because behind one mobile IP are thousands of real users, and anti-fraud systems cannot identify a specific bot pattern.

A practical tip for SMM specialists managing client accounts through Dolphin Anty or AdsPower: before linking a proxy to a profile, always check the Fraud Score, not just whether it "works or doesn't work." An IP may allow login but give a high Fraud Score, which could result in the account receiving reach limitations within 3-4 days.

Thresholds for marketplaces: Wildberries, Ozon, Avito

Marketplaces are less strict about Fraud Score than social networks because their main task is to prevent click fraud bots and return scammers, rather than to fight fake accounts as such. Benchmarks for thresholds:

Wildberries (seller's personal account, reviews, inflating) — a safe threshold is up to 20-30. Exceeding this increases the likelihood of blocking the personal account or rejecting reviews.
Ozon (multi-accounting for sellers) — acceptable up to 25-35, but the system monitors login patterns more closely than the IP itself.
Avito (posting ads from different cities) — up to 30-40 usually passes without problems, but it is more important that the geolocation of the IP matches the declared city of the ad.

For tasks where you need to appear as a "live" buyer or seller from a specific region, residential proxies are optimal — they provide IPs from real home internet users, consistently maintain a Fraud Score in the range of 5-20, and allow precise targeting to the desired city or region.

Thresholds for scraping and price monitoring

Here the situation is fundamentally different. When scraping prices on Wildberries, Ozon, or Yandex.Market, you are not creating an account and not performing actions on behalf of a user — you are simply reading public pages. Therefore, Fraud Score is important here not for "trust" but to ensure that the request does not fall under rate-limit or captcha.

Mass price scraping — an acceptable threshold is up to 40-50, as marketplace websites focus more on the frequency of requests than on the reputation of a specific IP.
Scraping with bypassing protection (Cloudflare, PerimeterX) — here the threshold is more important; it is better to keep it below 30, otherwise, the protection will require captcha on every tenth request.
SEO monitoring of search results (Google, Yandex) — recommended up to 20-25, because search engines aggressively ban IPs with high Fraud Scores from the results for several hours.

For clean scraping without logging into an account, it is often more advantageous to use data center proxies — they are cheaper and faster than residential ones, and their Fraud Score, although higher than average, is rarely critical for tasks where there is no strict anti-fraud check on user behavior.

How to check the Fraud Score of a proxy

Before uploading a proxy pool to a working project, it is worth checking each IP or at least a sample from the pool. The steps for a person without technical experience:

Open one of the open checking services — IPQualityScore, Scamalytics, or a similar one.
Copy the IP address of the proxy (it can be found in the settings of the anti-detect browser or in the personal account of the proxy provider).
Paste the IP into the checking field on the service's website and start the analysis.
Pay attention not only to the final Fraud Score number but also to the notes: "VPN", "Proxy", "Recent Abuse" — these flags also influence the platform's decision, even if the number itself is low.
For a pool of dozens of IPs, it makes sense to check 10-15% of the addresses selectively — if the majority's score is within the norm for your task, the pool can be considered operational.

Important: for residential and mobile proxies with dynamic IP rotation, the test result is relevant at the time of the request. If the provider changes IPs every 10-30 minutes, it makes sense to check the score periodically, rather than just once upon connection.

How to lower the Fraud Score: choosing the type of proxy

There is no single way to "lower" the Fraud Score of a specific IP — it is not a setting, but the reputation of the address, which is formed by the provider and usage history. However, there are practical ways to keep the score under control:

Choose the right type of proxy for the task — data center IPs almost always start with a higher base Fraud Score than residential or mobile ones, regardless of how carefully you used them.
Do not use one IP for dozens of accounts simultaneously — connection density is one of the key factors for increasing the score.
Check the pool before purchasing — quality providers offer a trial period specifically so that clients can run IPs through anti-fraud services.
Monitor the match between geolocation and time zone — discrepancies sharply increase the suspicion of the session regardless of the "cleanliness" of the IP itself.
Avoid free and public proxy lists — such addresses are flagged in dozens of blacklists and almost certainly have a Fraud Score above 70.

Comparison of proxy types by Fraud Score

Type of Proxy	Typical Fraud Score	Best suited for
Mobile Proxies	0-10	Facebook Ads, TikTok Ads, new account registration
Residential Proxies	5-20	Instagram, SMM multi-accounting, Wildberries, Ozon, Avito
Data Center Proxies	30-60	Mass price scraping, SEO monitoring without logging into an account

The scores are averaged and depend on the specific provider, but the general trend is stable: the closer the type of IP is to a "real person," the lower the base Fraud Score and the wider the range of tasks where it passes checks without additional barriers.

Conclusion and recommendations

Fraud Score is not an abstract metric but a working tool that should be checked before launching a campaign, not after the first wave of bans. There is no universal safe threshold: what is normal for price scraping on a marketplace will certainly lead to a ban in Facebook Ads already at the account registration stage.

A simple rule of thumb: for social networks and advertising accounts, keep the score below 15-20; for marketplaces, up to 30-40 is acceptable; and for pure scraping without authorization, you can work with a score of up to 50. If you are farming accounts for Instagram or TikTok, we recommend starting with mobile proxies — they consistently provide a minimal Fraud Score and reduce the risk of shadow banning at the start. For multi-accounting on marketplaces and precise geo-targeted work, residential proxies are better suited, while for mass data collection without logging into an account, more budget-friendly data center proxies are advisable.