Robots call after launching Yandex Direct — why this happens and how to protect your business

A company's entry into paid search and the Yandex Advertising Network (YAN) makes the business visible not only to buyers, but also to automated contact collection systems. However, in practice, many myths have accumulated around this topic: not every spam call means a deduction from the advertising budget, and far from every suspicious visit in the reports leads to financial loss. Let's analyze the mechanics of phone spam, the financial side of the issue, and engineering protection methods.

1. Launched Yandex Direct — and the phone started being attacked by robots

A familiar scenario for many advertisers: a campaign passes moderation, initial impressions begin, and within the first hours or days, the sales department faces a wave of non-targeted incoming calls.

Instead of real customers, managers hear the same thing on the phone:

• recorded voice clips offering 'to buy hot leads in your niche' or 'intercept visitor numbers from competitor sites';
• cold calls from robo-informers offering SEO promotion services, CRM implementation, or banking services;
• ringing silence for 20–40 seconds, after which the call is automatically disconnected;
• annoyed people who claim they just received a call from the company number, even though they did not leave any requests.

Why does business link this specifically to Yandex Direct? The correlation in time is too obvious: while the advertising account was paused, the line was quiet; as soon as the balance was topped up and ads were activated, a barrage of calls began. The owner's natural reaction is to assume that bots are purposefully clicking on their advertising, and the budget is being spent on paying for spam.

To sort out the problem, it is necessary to separate facts from assumptions and answer the three main questions of the investigation:

1. Does the advertiser really pay Yandex every time their number gets into an auto-dial database?
2. In what cases does a robot's call actually debit money from a Direct account?
3. What technical traces can be used to unambiguously prove the link between phone spam and advertising traffic?

2. How the advertiser's phone number ends up in auto-dial databases

For a company's number to end up in a telephone robot's queue, spammers do not always need to interact with the site like a normal user. The industry of automatic contact collection (scraping) works through several channels at once.

2.1. Monitoring commercial search results

Automated parsers work continuously on high-frequency commercial queries ('apartment renovation', 'dental implantation', 'cargo transportation', 'legal services', 'building materials supply'). They scan the Yandex search results and record the domains that appear in the special placement and guaranteed impressions blocks.

Then the script extracts the company phone number in one of three ways:

• Directly from the ad snippet — if the number is specified in the virtual business card of the ad or pulled from a linked organization profile;
• From a card in Yandex Business / Yandex Maps and other public directories, matching the advertiser's domain with public databases of legal entities and individual entrepreneurs;
• By parsing the site itself — by navigating to the landing page or the 'Contacts' section and reading links in the format from the HTML code tel: or sequences of digits by regular expression.

2.2. Why advertisers are interesting to B2B spammers

For B2B service providers, the fact of your presence in Yandex's paid results is the main qualifying trigger:

• Confirmed budget: if a company is running contextual advertising in a competitive niche, it means it is currently spending money on customer acquisition.
• Pain point: it is easier for an advertiser to sell 'cheap leads for 150 rubles', 'click-fraud protection', 'sales department training', 'CRM installation', or 'SEO instead of expensive Direct'.
• Full automation: modern lead generation chains work without human intervention — a parser finds a new domain in the search results, enriches it with a phone number, and instantly sends the contact to a cloud-based voice dialing system via API.

2.3. Is it necessary for a bot to click on an ad?

No, and this is a crucial technical nuance:

1. Number collection without visiting the site: a bot can see an ad in search, copy the domain from the snippet's address bar, and go to your site directly (direct traffic), or simply pull the phone number from the business listing in search. In this case, no advertising click occurs.
2. Click-through on a paid ad: more primitive parsers follow the advertising link directly in the search results to harvest the number from the landing page.
3. Difference between parsing and click fraud: the task of a parser is to get your phone number once and add it to a database. Even if it clicked the ad once and then your company received calls from 20 different robots from partner spam databases, this does not mean 20 clicked visits. Intentional clicking (click fraud), conversely, is aimed at repeatedly exhausting the budget or falsifying conversions. We wrote about basic traffic filters in detail in guide to combating ad click-fraud.

3. Four scenarios of automatic dialing

All automated calls that a business faces when working with contextual advertising fall into four scenarios with completely different technical natures and levels of financial risk.

Scenario #1. Commercial B2B auto-dialers

Collecting phone numbers of active advertisers for cold phone sales (informant robots offering leads, loans, marketing services).

• Threat: is annoying to employees, occupies the line, but in most cases does not cause direct damage to the advertising budget (the contact is usually taken without a series of paid clicks).

Scenario #2. Possible conversion fraud in YAN

The most toxic scenario for advertisers using auto-strategies with optimization or pay-per-call:

1. Bot performs a transition via an ad on a low-quality site in the YAN.
2. On the landing page, the dynamic call tracking script assigns a unique replacement number to this visit.
3. The attacker's script reads the substitute number and initiates an automatic phone call through the SIP gateway.
4. Background noise, silence, or a pre-recorded voice phrase lasting 35–45 seconds (slightly longer than the standard 30-second "targeted call" threshold in analytics systems) plays in the handset.
5. Call tracking records a 'target call' and sends the goal achievement to Yandex Metrica.

• Threat: under the 'Pay-per-conversion' model, the full cost of a lead is deducted from the balance, and the Direct autostrategy receives a false signal and starts buying even more junk traffic from fraudulent platforms.

Scenario #3. Attacks on callback forms (Callback fraud)

Bots visit the site and automatically fill out "We will call you back in 28 seconds" widgets, inputting real phone numbers of third parties into the input field.

• Threat: your phone system and managers automatically call back random people who did not place any orders. The company spends money on outgoing telephony and sales department work hours, receives negative feedback, and fake conversions for the 'Callback form submission' goal fly into Metrica.

Scenario #4. Intentional competitive attacks

A combination of automated ad clicks on expensive high-frequency keywords with simultaneous phone flooding (continuous dialing by robots to the company's main number).

• Threat: an attempt to simultaneously accelerate the burning of the campaign's daily limit and paralyze the multi-channel phone line so that real clients cannot get through during peak hours.

4. What the advertiser actually pays for

To understand if you are losing money on robot calls, you need to clearly distinguish between billing units in Yandex Direct and events in your telephony.

• Ad display: the mere presence of your ad on search does not debit money (with standard performance strategies). If a bot has scraped your domain or number from the results without clicking on the link — expenditure in Direct is 0 RUB.
• Click on the number in the ad (Virtual business card): if a user or bot clicks the call button directly in the Yandex ad block, the system charges it as a click. But the act of clicking the number on a smartphone merely opens the dialer screen and does not guarantee an actual call.
• Cost-per-click (CPC): money is deducted for each transition to the site that the Yandex anti-fraud system has deemed valid. If, after one valid transition, your number ended up in a spam database and 10 robots called you, you paid (possibly) for only one click, not for 10 calls.
• Cost-per-acquisition (CPA): payment depends on the selected goal and its link to the advertising interaction according to strategy terms. If the paid goal serves as "Click to call" or an unprocessed form submission, a false action can get into the account. The click itself on tel: does not prove a debit: check the source, goal achievement, filtering and expenditure in Direct. A long call also does not confirm lead quality — call tracking can transmit a robot as a target call if it only checks duration.

Cheat sheet: how to distinguish events and financial consequences

5. What Yandex officially says about bots, click fraud, and refunds

Compare suspicious visits with expenses in Direct. In official help on statistics Yandex explains how to match data and request a review.

5.1. How Direct anti-fraud works

• Filtering: Yandex describes checking invalid clicks online and re-analysis. Recognized invalid clicks are not paid for; after re-checking, a refund is possible.
• Practical check: a short visit in Webvisor does not in itself prove a debit. Compare it with the reports of the "Direct" group in Metrica and advertising campaign statistics.
• Data for investigation: the appearance of the visit and its duration do not allow reliably distinguishing a human from a bot. Compare the source, goals and confirmed inquiries.

5.2. What needs to be checked on the advertiser's side

• Call source: publishing a number and an incoming call do not yet establish a connection with paid advertising. Check the ad click, replacement number and visit identifiers.
• Goal quality: if call tracking transmits all long calls as target ones, a robot can also get into analytics. Check the content of the conversation and the processing result in CRM.
• Timeliness: save data when an anomaly is detected and contact support without delay. Clarify the available verification period and the composition of details for the specific request.

5.3. How to request a review of debits

In help on ad display the "Invalid traffic" report is described in the Report Library. It shows counted and filtered clicks; this is not a list of all refunds for each application.

In the request, specify the ad and campaign number, the period of the anomaly, the counter number and links to reports. To analyze calls, prepare comparable call tracking data: time, replacement number, visit identifier and qualification result. Transmit the necessary information through the provided support channel. Confirmation of invalid traffic and the decision on a refund is made by Yandex; the presence of spam calls alone does not guarantee a refund.

6. How to independently prove the link between calls and advertising

Before changing campaign settings or filing claims against the platform, conduct an internal technical audit in three directions.

6.1. Analysis of Yandex Direct and Metrica

1. Reconcile the timeline: correlate the date and time of the first spam calls with the moment campaigns were launched, bids were changed, or rankings reached the top of search results.
2. Check YAN platforms: in the Report Wizard, build a breakdown by the names of external networks and platforms (especially mobile apps and sites with an abnormally high CTR above 5–10% and minimal time on site).
3. Study the conversion report and Webvisor: check visits where phone goals were triggered. Pay attention to identical devices, direct traffic without scrolling to the footer or header, and instant tab closure exactly after 30–40 seconds.

6.2. Analysis of telephony and call tracking

1. Check which specific number the call is coming to: are robots calling your main landline or mobile number (which is listed in directories and the company profile), or are calls going strictly to a pool of dynamic call tracking numbers allocated for Yandex Direct?
2. Study SIP logs and Caller ID: spam networks often call from a pool of similar numbers (differing by the last digits) and have the same call duration (e.g., exactly 32 seconds of silence before hanging up).
3. Reconcile the call log with Metrica goals: did call tracking report these empty and robotic calls to Yandex Metrica as a conversion achievement?

6.3. Control experiment

If there is one general number on the site and you are not using call tracking, it is impossible to establish the source of the leak. To get an accurate answer, conduct a clean test:

• connect a separate virtual SIM card or city number from your operator that has never been published online before (not on Yandex Maps, nor on 2GIS, nor on social media);
• display this number exclusively for traffic from Yandex Direct (or in a specific advertising campaign);
• record all incoming calls to this number and compare them with the number of paid visits in Direct.

Practical incident investigation table

Use this matrix to classify each suspicious call and understand if your budget is suffering:

7. How to protect your business: five levels of technical defense

Since Yandex controls only the advertising click, and the phone call goes through communication networks, reliable protection is built on the principle of layered defense: from the website markup to the CRM system.

Level 1. Control of public phone numbers

• Separate phone numbers by channels and environments: never use the same phone number for Yandex Business (Maps), organic search, and paid campaigns in Direct. The organization's public number should remain available for existing customers and search engines, while paid traffic should be directed to dedicated dynamic pools.
• Remove the phone from the virtual business card at the start: if search parsers are working aggressively in your niche, do not expose the direct sales department number in the ad itself — drive traffic to a site where you can control visitor behavior.
• Caution when hiding phone numbers via JavaScript: we strongly advise against hiding phone numbers en masse using scripts (e.g., requiring a mandatory 'Show number' click or dynamically loading contacts based on mouse movement). This destroys the mobile user experience, creates insurmountable accessibility barriers for screen readers, and harms local SEO, where search engine crawlers expect valid Schema.org micro-markup. Protection should be built not by hiding contacts from real people, but through strict channel isolation and monitoring suspicious sessions at the analytics level.

Level 2. Website and callback form protection

• Abandon helpless callback widgets: any script that, after entering a number in the browser, instantly commands the telephone exchange to dial the subscriber is an open door for callback attacks.
• Implement server-side validation: connect invisible bot protection (Yandex SmartCaptcha or Cloudflare Turnstile) with strict token verification on the backend; add hidden honeypot-fields and time-to-fill form validation (a human cannot fill out and submit a form in 0.3 seconds after the DOM loads); set up strict Rate Limiting (limiting the number of requests from a single IP and session and prohibiting duplicate applications for the same phone number).
• Disable blind auto-dialing: switch callback requests to a manager preview mode or require number confirmation via verification code during a suspicious session.

Level 3. Call filtering at the telephony level

Even if a bot has parsed your number, it should not reach a live manager and trigger the target call timer:

• Enable anti-spam filters of your operator and cloud telephony: most virtual phone systems have regularly updated blacklists of known spam pools and SIP gateways.
• Implement a short IVR barrier (voice menu): if you are being attacked by silent robots or automatic informers, put a short greeting before connecting to the sales department: "Hello! You have reached the company. To speak to a specialist, press 1" (or set up voice recognition by an AI secretary).
• Why this works and how to assess the risk: typical auto-dialers and conversion fraud scripts perform calls on a rigid timer and do not handle DTMF tones in response to a voice menu. Incoming spam calls are cut off before connecting to an operator and are not counted as targeted calls. For real customers in urgent demand niches (tow trucks, emergency lock picking), an extra action creates unnecessary friction, so enable IVR selectively — during spikes of spam attacks or on dedicated lines from suspicious platforms.

Level 4. Correct call tracking setup

• Protect your pool of replacement numbers: configure the display of a dynamic number not immediately upon page opening, but with a minimal delay after checking the session for bot activity. If an obvious bot has visited the site, the call tracking script should not allocate a number from the advertising pool to it (display a default static placeholder number without linking it to Metrica).
• Reconsider the 'Target Call' criteria: increase the minimum conversation duration threshold from 30 to 45–60 seconds (and count it strictly from the moment a live operator answers, not from the beginning of the voice greeting!), enable tracking only unique (first-time) target calls and use tagging and speech analytics (recognition of silence, answering machines, and spam scripts) so that calls with silence are automatically marked as non-target and are not uploaded to Yandex Metrica.

Level 5. Protecting advertising strategy training

The most destructive effect of phone fraud is the breakdown of Yandex's auto-strategy algorithms. When a fake call hits a Metrica goal, the Direct algorithm considers the platform effective and directs your entire budget there.

• Disable optimization for micro-conversions: never set 'Click on phone number', 'Go to messenger', or 'Show number' goals as the main goals for autostrategies (especially with pay-per-conversion).
• Switch to transferring qualified leads from CRM (Offline conversions): set up the integration so that when a call or lead is made, a deal is created in the CRM with an attached ClientID Yandex Metrica; a manager (or speech analytics script) verifies the inquiry and changes its status to "Qualified lead" (real person, targeted request, required region); only after that does the CRM send an offline conversion to Yandex Metrica, which the strategy in Direct is trained on.
• Regularly clean YAN and cut off bot segments: exclude junk platforms and mobile apps, and also create segments of suspicious visits in Metrica and apply a negative bid adjustment to them -100% in Direct.

8. Legislation of the Russian Federation: how to fight illegal auto-dialing in 2026

In addition to technical blocks, a strict legal framework against telephone spam and auto-dialer robots has been established in Russia. If real Russian legal entities or individual entrepreneurs (lead sellers, marketing agencies, banks) call your company, their actions directly violate the law.

1. Direct ban on advertising auto-dialing (Federal Law "On Advertising")

According to Part 1 and Part 2 of Article 18 of the Federal Law No. 38-FZ 'On Advertising':

• dissemination of advertising over telephone networks is permitted only with the prior consent of the subscriber to receive such advertising. Moreover, the burden of proving the existence of such consent lies entirely with the advertiser (the fact that your number is published on a site or in Yandex Direct to receive orders from your clients is not consent to receive third-party B2B advertising!);
• the law directly prohibits the use of means for selecting and (or) dialing a subscriber number without human participation (automatic dialing, automatic mailing) for distributing advertising.

Violation of these rules is subject to administrative liability under part 4.1 of Article 14.3 of the Code of Administrative Offenses of the Russian Federation: the fine for individuals is from 10,000 to 20,000 rubles, for officials and individual entrepreneurs — from 20,000 to 100,000 rubles, and for legal entities — from 300,000 to 1,000,000 rubles. for every proven episode.

2. Regulation of mass calls (Article 44.1-1 of the Federal Law "On Communications")

In accordance with Article 44.1-1 of the Federal Law No. 126-FZ 'On Communications' (introduced by Federal Law dated 01.04.2025 No. 41-FZ), strict rules have been introduced for any mass and automatic telephone calls on the public communication network:

• Contract with the operator: the client of mass calls is obligated to conduct them exclusively based on a special contract with a telecom operator. Mass automatic calls bypassing this procedure are recognized as illegal.
• Right to opt-out via operator: a subscriber has the right to send a refusal to receive mass calls to their telecom operator, and the operator is obligated to stop forwarding such calls to the subscriber.

3. How to file a complaint with the FAS and telecom operators

If lead sellers or marketing service providers are systematically calling you:

1. Record the evidence: save the audio recording of the call from your telephone system, record the date, exact time, caller number (Caller ID), and your number that received the call.
2. Identify the beneficiary (if a robot calls without a company name): if a robot suggests 'press 1 to receive a commercial proposal in a messenger or connect to a manager', as part of control verification, you can wait for the transfer to a live manager or the arrival of a message to record the TIN, domain, or company details of the ordering client.
3. Request a detailed statement from your telecom operator for the day of the call, confirming the fact of an incoming connection.
4. Send a complaint: via a unified contact form for spam advertising on the website FAS Russia (by attaching an audio recording, call logs, and a written statement regarding the absence of your consent to advertising), as well as through your telecom operator's spam call reporting service to block the spammer's number at the operator's anti-fraud platform level.

Why won't the law eliminate 100% of telephone spam?

Legal mechanisms work perfectly against real Russian companies that order cold calling to sell their services. However, against anonymous technical fraud (inflating conversions in YAN via substitute SIP accounts or competitive phone flooding), a complaint to the antimonopoly service will not help — here, only the five-level technical protection described above saves you.

9. Checklist: is your advertising budget protected from phone fraud?

Go through these 10 control questions together with your marketer or advertising contractor. Each 'No' answer in the protection points is a potential hole in your budget:

1. Can your advertising number be obtained without visiting the site? How to do it correctly: the number in ads and directories is separated from the numbers on landing pages, or the business card is disabled in campaigns subject to parsing.
2. Is a separate phone number used specifically for the advertising channel? How to do it correctly: you know for sure that a call to this number could only have come from Yandex Direct, not from Yandex Maps, classifieds, or an old directory database.
3. Is it visible in analytics which specific campaign, Yandex Direct ad network site, and ClientID did a specific call come from? How to do it correctly: dynamic call tracking linked to Yandex Metrica and CRM is connected.
4. Are calls checked before the system considers them targeted? How to do it correctly: calls with silence, disconnections, and spam offers are filtered out by duration (from 45–60 seconds of pure conversation), uniqueness, or speech analytics.
5. Is the inclusion of spam calls in the goals used for training Direct auto-strategies excluded? How to do it correctly: autostrategies do not learn from 'Clicks on a number' or raw unfiltered calls.
6. Are anti-spam filtering and IVR protection configured on your telephony? How to do it correctly: known spam numbers are blocked automatically, and during attacks by silent robots, a tone filter ('Press 1') or voice assistant is activated.
7. Are callback forms protected against automated submission? How to do it correctly: all forms have SmartCaptcha or Turnstile validation on the server, request frequency limits, and disabled uncontrolled auto-dialing to unverified numbers.
8. Do you regularly check the 'Invalid traffic' report and YAN platform statistics? How to do it correctly: weekly track the share of filtered clicks and block platforms with signs of bot activity.
9. Are logs saved (ClientID, call recordings, timestamps) to dispute suspicious charges? How to do it correctly: when an anomaly is detected, the data related to it is saved and a request is sent to support; conditions and the available verification period are clarified for each specific case.
10. Is your advertising optimized for actually qualified leads from CRM? How to do it correctly: offline conversions are passed to Yandex Direct only after a manager has confirmed the adequacy and targeted nature of the inquiry.

10. Conclusion: why protecting just Yandex Direct is not enough

The problem of automatic calls when launching contextual advertising cannot be solved by one magic button in the advertising account for three fundamental reasons:

1. The advertising system's anti-fraud and telephony protection solve different tasks. Yandex algorithms check the technical purity of a click in the browser, but do not see what happens on your phone line after the number is read from the screen.
2. Contact collection is not always related to paid click fraud. Your phone number can be retrieved from search results or open sources for free, so panically disabling advertising at the first robot call often deprives a business of real customers.
3. Each suspicious call must be verified using a chain of data, not guessed. Only combining logs from telephony, call tracking, and Yandex Metrica shows the real picture: where a harmless cold parser just called you, and where a fraud network in YAN is currently debiting your budget for fake conversions.

Effective advertising budget protection in the modern environment is always a comprehensive engineering architecture where the website, analytics scripts, call tracking, virtual telephony, CRM system, and Yandex Direct autostrategies operate as a single filter. At STARCODE, you can order a comprehensive protecting ads from bots and clicking, deep advertising campaign audit or system Yandex Direct setup and management with an end-to-end integration of call tracking and CRM.

Sources and documentation

Sources verified October 26, 2026.

• Yandex Direct Help: protection against invalid clicks (anti-fraud) and traffic verification — Rules for automated filtering of invalid clicks, report reconciliation, and requesting reviews.
• Yandex Direct Help: "Invalid Traffic" report and impression verification. — Analysis of counted and filtered clicks in the Yandex Direct Report Library.
• Yandex Metrica Help: transmitting offline conversions and calls from CRM — Linking qualified inquiries and calls to ClientID for analytics and auto-strategies.
• Law No. 38-FZ 'On Advertising' — Articles 18, 18.1 and 24: newsletters, internet advertising and advertising of medical services.
• Code of the Russian Federation on Administrative Offenses (KoAP RF) — Part 4.1 of Article 14.3: administrative liability for violation of requirements for advertising via telecommunications networks.
• Federal Law No. 126-FZ 'On Communications' — Article 44.1-1: rules for making bulk calls and the subscriber's right to opt out of receiving them.

The documentation describes platform capabilities. We check configuration and compatibility for the specific project; STARCODE prices and SLA are determined by our proposal and contract.