How to Track Whether Your Ads Run in the Right Location

Ads can appear outside a target location because the campaign includes people interested in that location, the platform inferred location from imperfect signals, a device moved, a VPN or carrier gateway changed the apparent location, or reporting used a different geo field. The first task is to define what "in target" means, then verify delivered impressions against that rule.

The first question: did you target presence or interest?#

Many location disputes begin with a setting, not a delivery failure. A campaign targeting Delhi might be configured to reach people physically in or regularly in Delhi, plus people elsewhere who have shown interest in Delhi.

Google Ads describes two positive-location approaches in its advanced location options:

• Presence or interest: people likely to be in, regularly in, or interested in the selected location.
• Presence: people likely to be in or regularly in the selected location.

The broader option can produce legitimate impressions outside the physical boundary. It may be useful for travel, real estate, education, relocation, tourism, and other intent-led campaigns. It may be unsuitable when delivery must stay within a licensed, serviceable, regulated, or operational area.

Before calling the delivery wrong, export the exact positive and negative location settings used during the affected period. Settings can change mid-campaign, so the current screen may not represent the full reporting window.

What does an ad platform use to infer location?#

Platforms infer location from a combination of signals. Depending on the product and device, those can include:

• IP address and network routing;
• device location settings and permissions;
• mobile carrier information;
• account, search, and content behavior;
• recent or regular presence patterns;
• publisher or app signals; and
• location interest expressed through queries or viewed content.

No single signal is always available or correct. Google notes that location targeting is a best-effort process based on signals such as settings, devices, and behavior, and that 100 percent accuracy is not guaranteed.

The measurement team should keep at least three fields separate:

1. Targeted location: what the campaign was configured to buy.
2. Platform-reported user location: the location assigned inside the buying platform.
3. Independently observed location: the location inferred by a separate measurement path at delivery time.

These fields may disagree without any party intentionally misreporting. The value comes from measuring the pattern, identifying its cause, and deciding whether the residual fits the campaign's rules.

Seven common causes of ads outside the target area#

1. The campaign uses a broad location option

Presence-or-interest settings can deliberately include people outside the boundary. A user in Bengaluru searching for hotels in Goa may qualify for a Goa-targeted travel campaign even though their current physical presence is elsewhere.

Action: if physical presence is required, select the appropriate presence-focused option where the platform and campaign type support it. Record why the narrower choice is needed because it can reduce scale.

2. Targeting and observation were confused

In some campaign products, an audience or content segment can be added for observation without restricting delivery. Google's targeting and observation guidance distinguishes between narrowing reach and merely collecting reporting for a selected group.

Action: inspect whether each audience, placement, topic, and location control actually restricts reach. A label in the report is not proof that the setting acted as an exclusion.

3. IP geolocation disagrees with physical location

IP-based location is an inference about a network endpoint. Mobile carrier routing, corporate networks, satellite providers, shared gateways, and stale database records can place the apparent IP in another city or state.

Action: retain the provider name, lookup timestamp, database version, confidence or precision field, and raw classification used for the decision. Do not treat city-level IP output as exact GPS evidence.

4. VPNs, proxies, and privacy relays changed the apparent location

A legitimate person can appear in another region because of a corporate VPN, consumer VPN, privacy relay, or security gateway. Invalid operators can also use proxies to imitate in-target traffic.

Action: combine the geo result with ASN, hosting, proxy, device, and behavior signals. A location mismatch alone does not prove fraud. A mismatch plus data-center routing and automated behavior is a stronger investigation lead.

5. The user or device moved

A person can click after traveling, return to a site later, or use a device whose regular-location model lags behind current presence. Campaign reports and analytics may therefore assign different locations to related events.

Action: compare event timestamps and avoid joining a later conversion location back to an earlier impression as if they were measured simultaneously.

6. The report uses billing, account, or publisher location

A field labeled "country" may refer to the user's inferred location, publisher market, account setting, billing country, content locale, or data center. Comparing unlike geo dimensions creates false leakage.

Action: obtain the report schema. Write a plain-language definition for every geographic field and confirm which event supplied it.

7. The inventory or event was misrepresented

Some mismatches do reflect poor-quality or invalid delivery. App spoofing, domain misrepresentation, falsified measurement events, manipulated device signals, and invalid proxy traffic are recognized IVT concerns. The MRC's IVT standards addendum provides the relevant taxonomy.

Action: investigate at source, placement, app, device, ASN, creative, and time-window level. Avoid applying a fraud label to every out-of-area observation.

Why platform and analytics geo reports disagree#

The ad platform may infer a location at impression or click time. GA4 may infer location when the website event arrives. An independent verifier may enrich the impression from another provider. The CRM may contain a declared address entered days later.

Each field answers a different question:

Do not select one column as universally true. Reconcile the event, timestamp, precision, and method.

A practical geo-verification workflow#

Step 1: write the business rule

Replace "target Maharashtra" with a testable statement:

Count an impression as in target when the delivery-time location resolves to Maharashtra at state-level precision. Report city-level data for analysis, but do not reject an impression solely because a city cannot be resolved.

Your rule may instead use country, designated market area, postal code, radius, or a set of serviceable regions. It should also state how to handle border areas and unknown locations.

Step 2: preserve the campaign configuration

Save:

• included locations;
• excluded locations;
• presence or interest option;
• audience expansion or automated targeting settings;
• campaign and ad-group overrides;
• start and effective times for every change; and
• the platform's location report for the same period.

Screenshots help humans, but structured exports are better for reconciliation.

Step 3: collect delivery-time evidence

For each measured impression, retain the identifiers and fields needed to reproduce the classification:

• event ID and timestamp;
• campaign, creative, source, placement, and app or domain;
• channel and device class;
• IP-derived country, region, and city where permitted;
• ASN, network, hosting, VPN, proxy, or relay classifications where available;
• geo provider and database version;
• target rule result; and
• fraud or quality flags kept separate from geo status.

Apply privacy minimization, access control, and retention rules. Most investigations need a derived location and network classification, not indefinite retention of a raw IP address.

Step 4: report four buckets

Avoid a single target-match percentage. Use:

1. Verified in target
2. Verified out of target
3. Unknown or insufficient precision
4. Excluded from quality reporting, such as invalid traffic under the agreed methodology

This prevents unknown data from being quietly counted as either compliant or non-compliant.

Step 5: find concentration

Break out the out-of-target and unknown rates by source, app, site, exchange, placement, device, ASN, hour, creative, and campaign setting. Concentration creates an actionable lead.

Examples:

• one carrier may have a regional routing issue;
• one publisher may be supplying a neighboring market;
• one source may show data-center traffic and impossible device patterns;
• a settings change may align exactly with the first leakage spike; or
• one geo provider may have stale classification for a recurring IP range.

Step 6: run controlled tests

Use approved test devices or partners in target and out of target. Record the campaign eligibility settings, observed ad behavior, public IP classification, device permissions, and timestamp. Do not rely on ad-preview tools as proof of production delivery. They answer whether an ad may be eligible under a test context, not what every real impression did.

Step 7: choose the response

The correct action depends on the cause:

• narrow presence settings;
• add explicit exclusions;
• remove or down-rank a source;
• correct the target rule;
• update a geo provider or classification;
• allow a documented carrier exception;
• separate unknown from non-compliant; or
• escalate evidence to the publisher or platform.

How do you calculate geo target match rate?#

Define the denominator before calculating the rate.

A transparent valid-traffic version is:

geo target match rate = verified in-target impressions / impressions with sufficient geo precision

Also report coverage:

geo coverage rate = impressions with sufficient geo precision / valid measured impressions

Suppose a campaign has:

• 1,000,000 valid measured impressions;
• 900,000 with sufficient state-level precision;
• 855,000 verified in target; and
• 45,000 verified out of target.

The target match rate among classifiable impressions is 95 percent. Geo coverage is 90 percent. Reporting only 95 percent hides that 10 percent of valid measured impressions could not be classified at the required precision.

Keep invalid-traffic counts visible in a separate quality section. Removing them from the target-match denominator may be appropriate under the measurement contract, but the report should disclose the removal.

Is an out-of-target impression always wasted spend?#

No. It depends on the campaign rule and why the impression was out of target.

An out-of-area traveler interested in the product may be valuable for a broad travel campaign. The same impression may be unacceptable for a regulated local offer or a service available only within one state. A location mismatch can also be a measurement artifact rather than actual delivery outside the area.

Classify the business impact separately from the technical result:

• in target and eligible;
• out of target but allowed by campaign policy;
• out of target and non-compliant;
• unknown, requiring no decision or follow-up under the contract; and
• invalid traffic, handled under the IVT policy.

What evidence should you send in a location-discrepancy ticket?#

Provide a bounded, reproducible package:

• campaign and line-item IDs;
• exact UTC time window;
• targeted and excluded location exports;
• presence or interest settings;
• count and rate for in-target, out-of-target, and unknown buckets;
• the precision required by the campaign;
• breakdown by source, device, and network;
• a sample of event IDs and timestamps;
• independent geo provider and database version;
• relevant VPN, proxy, hosting, and IVT classifications; and
• test results from known regions.

Share the minimum data required and follow the applicable privacy and contract controls. A list of raw IP addresses sent by email is rarely the right dispute process.

How AdProof supports target validation#

AdProof's measurement pipeline normalizes supported web, mobile, video, and VAST delivery events, then enriches them with geo, device, ASN, and IP-intelligence signals. Campaign target rules can be evaluated against those measured events, while fraud and source-quality results remain available as separate dimensions.

That gives a media team an independent view of where measured delivery appeared to occur and where mismatches concentrate. It does not turn inferred geo into perfect physical-location proof. The report should still disclose coverage, precision, provider, unknowns, and the campaign rule.

Explore AdProof target validation and delivery verification, review how invalid traffic is classified, or request a pilot for a campaign with material geographic constraints.

Location is only one part of campaign quality. Use the companion guides to check whether ads reached the right audience and build a brand-safety workflow.

Frequently asked questions#

Why are my Google Ads showing in other countries?

Check whether the campaign uses presence-or-interest targeting, whether nearby or related areas are eligible, and which location field the report displays. Then compare delivery-time geo evidence with the actual campaign rule. A foreign observation can also come from VPNs, network routing, travel, or geo-inference error.

Why are my Facebook or Instagram ads showing in other states?

Possible causes include broad audience or location settings, automatic expansion, platform location inference, device movement, network routing, and a reporting-field mismatch. Export the effective campaign settings and investigate delivered impressions by source and network before assuming fraud.

Can a VPN make an ad appear outside the target location?

Yes. A VPN or relay can change the location inferred from the network address. It can affect legitimate users and invalid operators, so combine the result with device, ASN, hosting, behavior, and other quality evidence.

Is IP geolocation accurate enough for city targeting?

IP geolocation can support country and regional analysis, but city-level precision varies by network and provider. Treat it as an inference, retain precision and coverage fields, and avoid presenting it as GPS proof.

Should unknown location count as out of target?

Only if the measurement contract explicitly defines it that way. The clearer default is to report unknown separately so stakeholders can see both target-match quality and measurement coverage.

← Back to all field notes