# Lead generation fraud in real estate: how to filter out fake leads

Author: Ruslan Matveev · Date: 2026-07-03 · Source: https://matveo.com/en/blog/frod-v-lidogeneratsii-nedvizhimosti · Russian original: https://matveo.com/blog/frod-v-lidogeneratsii-nedvizhimosti

Lead generation fraud in real estate: form bots, click fraud, incentivised calls. Signs of fake leads in your CRM and how deep CRM goals protect campaigns.

## How much fraud there is in new-build leads

According to estimates from [call tracking](https://matveo.com/en/blog/kolltreking-i-rechevaya-analitika) services and agencies, the share of fraudulent and non-target enquiries in new-build lead generation in 2026 ranges from 10 to 15 percent on Yandex search to 30 to 50 percent in overheated Yandex Advertising Network auctions and at some CPA networks. The exact figure depends on the region, the housing class and whether the project has any protection at all. The ranges are wide because fraud is distributed unevenly: one project runs almost clean, while the neighbouring one, on the same budgets, collects junk in a third of its enquiries. Nobody keeps honest market-wide statistics: neither the ad platforms nor the contractors have any interest in publishing them. Each team learns its own figure only when it first breaks enquiries down by source and calls them one by one. That moment usually comes after an unpleasant conversation about why the growing number of leads is not producing more meetings and reservations.

Real estate attracts fraud through its economics. A targeted new-build enquiry costs 1,500 to 20,000 ₽ depending on the segment, which makes it one of the most expensive enquiries on the Russian-language internet. Where an enquiry costs as much as an average restaurant bill, faking enquiries becomes a business: bot operators sell automated traffic, incentivised performers leave contact details for a reward, and dishonest contractors top up the plan from old databases. The higher the auction bids and the stricter the KPI on lead numbers, the more people want to sell the developer the appearance of demand. Add the long deal cycle: months pass between enquiry and contract, and in that time the source of junk manages to get paid and disappear. So the question is a practical one: how much junk is in your enquiries specifically, through which doors it gets in, and who on your side is responsible for noticing it.

## Bots fill in forms, people call for money: types of fraud

There are four main types of fraud in real estate [lead generation](https://matveo.com/en/blog/lidogeneratsiya-dlya-zastroyshchika): bots that fill in forms on landing pages and quizzes with plausible phone numbers; incentivised traffic, that is, real people leaving enquiries for a reward; click fraud on Yandex Direct ads; and the resale of other people's or old databases passed off as fresh leads. The types differ in the funnel stage where they can be caught, so the protection against each one is different too. An incentivised performer, following instructions, will even answer the first call and act interested in buying, while click fraud involves no enquiries at all and simply burns through the campaigns' daily budget.

Bots have become noticeably smarter in recent years. A modern script emulates mouse movement and typing pauses, passes simple checks, and takes numbers from generated ranges or leaked databases, so at first glance the enquiry is indistinguishable from a real one. Incentivised traffic is more insidious: the person confirms interest on the first call as instructed, and such an enquiry passes call centre qualification. It is exposed only deeper in the funnel, when the "buyer" stops answering the phone before the viewing. Click fraud in Yandex Direct most often hits expensive competitive queries, where a click costs hundreds of roubles: a competitor or a botnet only needs to drain your daily budget by lunchtime to take the search results for itself. Resold databases give themselves away differently: the person picks up and is genuinely surprised, because they were looking for a flat six months ago and have already bought one, or they left an enquiry for a different property.

## Where fake leads come from: CPA networks and volume guarantees

A noticeable share of fake new-build enquiries comes through intermediaries: CPA networks and lead generators that promise a fixed volume of leads at a fixed price. When a contractor is paid for every enquiry and bound by a volume guarantee, they have a direct motive to make up the shortfall by any means: incentivised traffic, bots or other people's databases. A guarantee that looks like protection for the client actually creates a conveyor belt of junk: the lead plan is always met, regardless of whether there are people with money and an intention to buy a flat behind those leads. [Marketing department](https://matveo.com/en/blog/otdel-marketinga-developera)s where the bonus is tied to the number of enquiries are especially vulnerable: there the junk flow suits both sides until sales asks about meetings.

The pay-per-lead model itself works; the question is control. An honest contractor discloses traffic sources, shows the campaigns and agrees to be paid per qualified lead after a call. Three things should raise concern: a suspiciously even flow of enquiries day after day with no seasonality or dips, a refusal to show placements and creatives, and a price noticeably below the market. If a qualified enquiry in your segment costs 8,000 ₽ and you are offered them at 2,000 ₽, the difference will almost certainly be made up in quality. There is one test question, and it saves months: is the contractor willing to tie payment to a meaningful call or a completed meeting instead of a submitted form. Someone selling real buyers agrees to such a scheme calmly. Someone selling rows in a spreadsheet starts haggling over the definition of a meaningful call.

## How fraud inflates CPL and breaks campaign optimisation

Fraud hits campaigns twice: first it eats the budget directly, because every fake enquiry is paid for with real clicks costing hundreds of roubles, and then it corrupts the data that automated strategies learn from, so the campaign starts drifting towards junk sources on its own. If the optimisation goal is a form submission and a third of submissions are made by bots, the algorithm considers the fraudulent placements the most productive and pours even more money into them. The advertising literally learns to bring in junk, and does so diligently: from the machine's point of view it is meeting the goal it was given.

Then a vicious circle starts. The CPL in the report looks decent, the director is calm, and the cost of a real deal quietly grows, because the share of junk in enquiries increases with every learning cycle of the strategy. Honest placements suffer too: the budget the algorithm shifted to fraudulent sources was taken from those that brought buyers. A quarter later the channel picture in the reports has little to do with reality, and media split decisions are made on distorted figures: the live channel is cut back for being expensive, the fraudulent one is scaled for being cheap. We have already covered why the average [cost per lead](https://matveo.com/en/blog/stoimost-lida-v-nedvizhimosti) is misleading even on clean traffic: fraud amplifies every flaw of the average figure and adds a systematic skew towards the worst sources.

## Signs of fraud in the CRM: how to check lead quality

Lead quality can be checked without special tools, straight in the CRM. The main signs of fraud: series of phone numbers from one range or with sequential digits, spikes of enquiries at night and at hours untypical for your audience, a 100 percent missed-call rate on a particular campaign or placement, identical names and template comments in the form fields. Each sign on its own can be a coincidence, but two or three signs on a single source add up to a diagnosis.

A normal missed-call rate on live traffic stays around 20 to 30 percent. If for some source it is consistently above 60 to 70, the source is almost certainly dirty. Speed is also worth looking at: an enquiry five seconds after the click means autofill, since a real person does not even have time to read the offer in that window. A practical technique: export a month of enquiries, break them down by Yandex Advertising Network placements and compare the connected-call share. Usually 80 percent of the junk is concentrated on 10 to 20 placements, and switching them off immediately makes the campaign healthier. This breakdown takes one evening and regularly saves a double-digit share of the budget. It is important to repeat it at least once a month: fraudulent placements in the networks keep changing, and a clean-up from a year ago no longer protects against fresh junk.

## Fraud protection: from phone validation to deep goals

Working fraud protection is built in four layers: an invisible captcha or behavioural scoring on the form against simple bots, phone validation via an HLR lookup, call tracking anti-fraud against incentivised calls, and campaign optimisation on deep CRM goals instead of form submissions. No single layer solves the problem on its own, but together they close all four types of fraud, each on its own floor of the funnel. Captcha and scoring barely cut the conversion of live visitors, number validation via an HLR lookup checks that the number exists and is active and costs pennies per enquiry, and call tracking anti-fraud flags suspicious calls by duration, source and the content of the conversation.

Deep goals solve the problem at the level of fraud economics. It costs a bot nothing to submit a form, and an incentivised performer can easily answer the first call, but imitating a two-minute meaningful conversation, a confirmed budget and a completed viewing is already expensive, and at scale almost impossible. When the automated strategy learns from mid-funnel events, fraudulent placements stop looking successful and lose budget automatically, without manual clean-up. This is the same closed-loop principle for advertising and analytics that we wrote about separately: the signal from sales flows back into the campaigns and cleans them out. In our practice, betting on the targeted enquiry instead of the cheap contact across four markets produced a conversion to sale of around ten percent with a fivefold return on traffic: signal purity turned out to matter more than enquiry volume. There were fewer enquiries, but almost one in ten turned into a contract, and the sales team stopped wasting time on empty call rounds.

## How an AI agent checks every lead and every source

An AI agent closes what a person cannot handle by volume: scoring every incoming lead and a daily reconciliation of traffic quality by source. A language model assesses the enquiry on a combination of signals: source and placement, the time and speed of filling in the form, the number pattern, on-site behaviour, the transcript of the first call. The output is a probability estimate that you are dealing with a buyer, and it lands in the CRM before the manager has picked up the lead. The difference from classic anti-fraud filters is that the model looks at a combination of weak signals: each on its own proves nothing, but together they detect junk more reliably than any single rule.

The agent's second loop is the daily reconciliation: every morning it compiles, by source, the connected-call share, the qualification rate and movement towards a deal, and highlights anomalies. This campaign's missed-call rate rose from 25 to 70 percent in three days, this placement has ten enquiries with numbers from one range, this ad gets all its enquiries between two and four in the morning. A person needs a day of work a week for such a breakdown; the agent does it before the working day begins. At Matveo we start with a two-week marketing audit: we break your traffic down by source, show the real share of fraud and non-target enquiries and what it does to the economics, and the cost of the audit is credited towards a pilot. In the case of a premium residential complex in St Petersburg it was precisely honest data on lead quality per channel that allowed us to keep a channel that looked loss-making on the top-line metrics: over the long window it paid back sixteen times over. Without a breakdown of enquiry quality that channel would have been closed after the first month, and the budget would have gone where enquiries were cheaper and deader.

## Frequently asked questions

Question: How can I quickly check whether there is fraud in my enquiries? Answer: Export a month of enquiries from the CRM and calculate the missed-call share for each source and campaign. Live traffic gives a missed-call rate of around 20 to 30 percent; for a dirty source it climbs towards 60 to 100. Additionally, look at phone number patterns, the time enquiries appear and the speed of form completion after the click. Such an audit takes an evening and requires no tools other than a spreadsheet.

Question: Does a captcha help against fake enquiries? Answer: Against simple bots, yes: an invisible captcha with behavioural scoring cuts off a noticeable part of automated junk and barely reduces form conversion. Against incentivised traffic a captcha is powerless, because the enquiry is left by a real person. That is why it is supplemented with number validation, call tracking anti-fraud and optimisation on deep goals.

Question: What share of non-target leads is considered normal? Answer: Completely clean traffic does not exist in real estate: even in healthy campaigns 20 to 30 percent of enquiries are screened out at qualification. The alarm threshold starts at roughly half. If more than 50 percent of a source's enquiries fail qualification, look into the placements, the keywords and the goal settings.

Question: What should I do about click fraud in Yandex Direct? Answer: Yandex automatically filters out some invalid clicks and refunds the money for them, but it catches far from everything. On your side, restrict impressions on suspicious Yandex Advertising Network placements, watch for abnormal spikes in clicks on expensive queries and file complaints with statistics exports: if confirmed, Yandex recalculates the charges.

Question: Can fraud be eliminated completely? Answer: Not completely; the task comes down to economics: making deception cost more than it earns the fraudster. The combination of form protection, number validation, anti-fraud in call tracking and optimisation on CRM events reduces the share of junk to a background level at which it stops affecting budget decisions and campaign learning. After that all you need is regular quality control by source, so that new schemes do not take root unnoticed.