Google indexes pages with the gtm_latency=1 parameter, and it shouldn’t!

The original article was written in Polish on January 7, 2026.

It all started with one line in the server logs.

While reviewing Googlebot traffic on one of the analyzed sites, I noticed something strange. The bot was visiting URLs with a ?gtm_latency=1 parameter that wasn’t anywhere in the site’s structure:

• I don’t have this parameter in internal linking
• No sitemap contained it (or contains it)
• The ?gtm_latency=1 parameter literally appeared out of nowhere, or rather was added by Googlebot itself, which I found out later

Googlebot first visited an address with this parameter on August 15, 2025 at 01:49.

A quick Google search yielded no answers. There’s no information about it in the Google Tag Manager documentation. SEO forums don’t know about it. I asked SEO specialists I know but they don’t know the parameter either. On exactly August 18, 2025, I asked the LinkedIn community… well, there was an attempt to help but I was getting similar claims from AI, which didn’t satisfy me at all because it falls under hallucinations.

Then on August 21, 2025, I wrote a new post on LinkedIn explaining what the ?gtml_latency=1 parameter is with some technical information and a protip not to block this parameter in robots.txt

I waited a month and on exactly Sep 22, 2025 at 13:48:03 a new publication appeared on LinkedIn, this time in English, in which I directly tagged personalities like Barry Schwartz and John Mueller asking about this parameter because Google had already indexed over 166k pages.

I asked:

• What’s behind the growing indexing of pages with the gtm_latency parameter?
• Two years ago, Gary Illyes mentioned that GoogleOther was meant to relieve the main index of tasks not related to search. Has anything changed in how GoogleOther affects search results?

However, I didn’t get an answer, and my post was… ignored 🥲

Why should every SEO specialist care about this?

If you work in SEO, you know how important it is to control what gets into the index. Duplicate content, cannibalization, crawl budget, render budget are the foundations of technical SEO, and here we have a situation where:

1. Google itself generates URL variants by adding diagnostic parameters
2. These variants get into the index even though they shouldn’t
3. Pages without proper canonical can have a real problem with duplication
4. Nobody’s talking about it!

So I’m starting 🙂

What is ?gtm_latency=1

I dug through probably half the internet looking for an explanation and found literally nothing. The topic is practically absent from the public space. A few people asked similar questions, but without answers. The official Google Tag Manager documentation is silent on this. You won’t find it in any guide, FAQ, or help article. I decided to look at the gtm.js source code – this is the file that Google Tag Manager loads on every page with an installed container.

What is the ?gtm_latency=1 parameter

?gtm_latency=1 is an internal diagnostic parameter of Google Tag Manager that activates extended collection of telemetric data about the performance and behavior of tags on the page.

Simply put: it’s a switch that tells GTM „enable full monitoring mode”.

function il(a) {
    var b = String(a[mf.Na] || "").replace(/_/g, "");
    return Kb(b, "cvt") ? "cvt" : b
}

var jl = x.location.search.indexOf("?gtm_latency=") >= 0 || 
         x.location.search.indexOf("&gtm_latency=") >= 0;

var kl = Math.random(),
    ll, ml = dj(27);
ll = jl || kl < ml;

var nl, ol = dj(42);
nl = jl || kl >= 1 - ol;

var pl = function(a) {
    pl[" "](a);
    return a
};
pl[" "] = function() {};

var kl = Math.random();      // random value 0-1
var ml = dj(27);             // = 0.005 (i.e. 0.5%)
ll = jl || kl < ml;          // true if drawn < 0.005

https://www.googletagmanager.com/td?id=...&v=3&t=t&pid=...

HTTP/3 204 
date: Thu, 21 Aug 2025 08:47:40 GMT
pragma: no-cache
expires: Fri, 01 Jan 1990 00:00:00 GMT
cache-control: no-cache, no-store, must-revalidate
content-type: text/plain
cross-origin-resource-policy: cross-origin
server: Golfe2
content-length: 0

What Google knows thanks to GTM telemetry

By collecting data from diagnostic mode, Google gets a complete picture:

About the site:

• URL and structure
• GTM container version and configuration
• Active experiments

About performance:

• Script loading times
• Resource loading sequence
• GTM position relative to other scripts

About tags:

• List of all tags in the container
• Execution status of each tag (success/failure/error)
• Execution times
• Dependencies between tags

About consents:

• Consent Mode state
• GDPR/RODO and DMA compliance
• Consent change history
• Impact of consents on tag execution

About errors:

• Which tags report errors
• Types of errors (exception, failure)
• Error context

The question is: who and why adds this parameter to URLs during crawling?

GoogleOther – the bot that was supposed to be invisible

In 2023, Gary Illyes from Google officially announced the introduction of a new user-agent: GoogleOther.

Its task was to relieve the main Googlebot from tasks not directly related to the search engine. GoogleOther is a bot designed for research and development (R&D) purposes – testing, experimenting, collecting diagnostic data.

Gary Illyes on GoogleOther

We’ve added a new bot, GoogleOther, to our list of bots, which will eventually relieve Googlebot. It’s a useless change for you, but I still think it’s interesting.

By optimizing how and to what extent Googlebot indexes, we wanted to ensure that Googlebot’s indexing tasks are used exclusively internally to build the index used by the search engine. To this end, we added a new indexing bot, GoogleOther, which will replace some of Googlebot’s other tasks, such as indexing for R&D work, to free up some indexing resources for Googlebot.

The new bot uses the same infrastructure as Googlebot, and thus has the same limitations and features: host load limitations, robotstxt (though a different user agent token), HTTP protocol version, download size, and so on. It’s basically Googlebot under a different name.

On April 20, 2023, Gary Illyes wrote this on his LinkedIn profile

It’s a logical solution, since if Google wants to test pages, check performance, conduct technical audits – this shouldn’t affect what users see in search results.

The theory sounds reasonable. Practice looks different.

Data from server logs

Analyzing server logs for pages where I noticed the ?gtm_latency=1 parameter, I identified the following patterns:

Observation:

The ?gtm_latency=1 parameter is added by the GoogleOther bot – exactly the one that according to Google was supposed to be separated from the main index.

The R&D bot performs a GTM diagnostic test. This fits its purpose.

The problem appears in the next step.

URLs leak into the main index

I checked how many pages with the ?gtm_latency=1 parameter are in the Google index:

inurl:gtm_latency=1

Growth timeline:

Half a million URLs with a diagnostic parameter in Google’s main index.

This shouldn’t happen in my opinion 🙂

Is this a bug?

Unfortunately, there is no official Google statement on this matter. I tried to get an answer:

• I tagged Barry Schwartz on LinkedIn – no response
• I tagged John Mueller on LinkedIn – no response
• I searched official Google channels – no mention of the problem

I can only speculate based on observations:

Hypothesis 1: This is an unintended bug

GoogleOther was supposed to be isolated, but somewhere in the indexing pipeline there’s a „leak” – URLs visited by GoogleOther end up in the main index contrary to assumptions.

Hypothesis 2: Policy change

Google changed how GoogleOther works but didn’t communicate it publicly. URLs are now intentionally indexed.

Hypothesis 3: Partial isolation

GoogleOther is separated only for some types of crawls but not for all. Crawls with the ?gtm_latency=1 parameter slip through the system.

Without an official Google statement, we remain in the realm of speculation.

Potential SEO consequences

Regardless of the cause, the fact remains: over half a million URLs with a diagnostic parameter are in the Google index. What does this mean for site owners?

Problem 1: Content duplication

Every URL with ?gtm_latency=1 is a potential duplicate of the page without the parameter:

site.com/product/name/                ← original
site.com/product/name/?gtm_latency=1  ← duplicate in index

If the site doesn’t have a properly configured rel="canonical" tag, Google may:

• Treat both versions as separate pages
• Split ranking signals between them
• Choose the „wrong” version as canonical (small chances but probable)

Problem 2: Keyword cannibalization

Two versions of the same page can compete for the same keywords:

Problem 3: Crawl budget

Every URL that Google indexes is a URL it must crawl. For large sites:

• 10,000 product pages = potentially 10,000 additional URLs with the parameter
• Doubling the number of URLs to crawl
• Slower discovery of new content
• Slower refresh of existing pages
• Worse result related to render budget

Problem 4: Report pollution

In Google Search Console there may appear:

• Pages marked as „Indexed, not submitted in sitemap”
• Discrepancies in data between URL versions

Note: In my observations, pages with ?gtm_latency=1 DID NOT appear in GSC – they were only visible in server logs. This may mean that Google treats them differently, or that the problem is not yet fully visible in standard tools, or that only results directly from the main Googlebot go to GSC, and GoogleOther has no connection with GSC.

Problem 5: Users land on diagnostic URLs

If a URL with the parameter is indexed, it may appear in search results. The user clicks and lands on a page with the parameter, and then:

• Shares the URL with the parameter
• Links to the URL with the parameter
• Bookmarks the URL with the parameter

The parameter spreads and not through direct Google action!

Canonical analysis – what the data showed

I conducted an audit of 142 randomly selected addresses with the ?gtm_latency=1 parameter visible in the Google index:

Interpretation:

• 77.5% of pages are „safe” – even if the URL with the parameter is in the index, Google should respect the canonical and show the correct version in results
• 10.6% of pages have no protection – lack of canonical means Google decides itself which version is „canonical”
• 6.3% of pages have a serious problem – canonical points to URL with parameter, which may perpetuate duplication
• 0.7% have incorrect configuration – canonical to homepage instead of the proper subpage
• 4.9% return errors – which may generate additional indexing problems

Recommendations and protips, my solutions to this problem

Since I now know the mechanism behind ?gtm_latency=1, the scale of the problem and potential SEO consequences, it’s time to answer the most important question: what to do about it?

Below I present specific recommendations based on my observations and analysis of the problem.

✅ Recommendation 1: Conduct a canonical tags audit

This is the absolute foundation of technical SEO hygiene. I have several scenarios:

Scenario A – Canonical configured correctly:

<!-- On page: site.com/product/name/?gtm_latency=1 -->
<link rel="canonical" href="hxxps://site.com/product/name/" />

Google sees canonical pointing to version without parameter. In most cases:

• Respects the canonical setting
• Will display the version without parameter in search results
• Will consolidate ranking signals on the proper URL

Scenario B – No canonical:

<!-- On page: site.com/product/name/?gtm_latency=1 -->
<!-- No canonical tag -->

Google doesn’t receive a signal about the canonical version. It must decide itself:

• Which version is „better”
• Which to display in results
• How to split ranking signals

In practice, it may choose the version with the parameter as canonical. Especially if it was indexed first!

Scenario C – Canonical points to itself (self-referencing with parameter):

<!-- On page: site.com/product/name/?gtm_latency=1 -->
<link rel="canonical" href="hxxps://site.com/product/name/?gtm_latency=1" />

This is the worst scenario. Google receives a clear signal: „Version with parameter is the canonical version” – this perpetuates duplication

How to conduct a canonical audit

Step 1: Check if your site is affected by the problem

Search in Google:

site:yourdomain.com inurl:gtm_latency=1

If the results are empty, with high probability the site doesn’t have indexed URLs with this parameter at the moment.

If results appear – you have confirmation of the problem and a list of specific URLs to check.

Step 2: For each URL found, check the canonical

Canonical should point to URL without the ?gtm_latency=1 parameter.

Problematic states:

<!-- PROBLEM: No canonical tag -->
<!-- (nothing found) -->

<!-- PROBLEM: Canonical with parameter -->
<link rel="canonical" href="https://yourdomain.com/path/to/page/?gtm_latency=1" />

<!-- PROBLEM: Canonical to homepage instead of proper subpage -->
<link rel="canonical" href="https://yourdomain.com/" />

Step 3: Check canonical implementation across the entire site

The problem may concern not only pages already indexed with the parameter but the entire logic of generating canonical tags.

Typical implementation errors:

❌ Recommendation 2: DO NOT block the parameter in robots.txt

This is the wrong approach. Don’t do it, here’s why I think so.

Reason 1: robots.txt blocks crawling, not indexing

This is a difference that many SEO specialists don’t fully understand:

1. robots.txt → Tells bots: „Don’t visit this page”
2. Bot doesn’t crawl the page but…
3. …if the page is already in the index or has external links…
4. …Google MAY continue to display it in results (without snippet)

By blocking ?gtm_latency=1 in robots.txt:

• Googlebot won’t crawl those URLs
• But if the URL is already in the index – it will remain in the index, because Google doesn’t see it shouldn’t be indexed
• If the URL has external links – Google may display it with the annotation „Description unavailable due to robots.txt”
• You won’t solve the duplication problem, only make it worse

Reason 2: You’re blocking access to diagnostic data

The ?gtm_latency=1 parameter serves Google to collect data about GTM performance. By blocking it:

• Google can’t conduct a performance audit of the site
• You lose potential benefits from GTM optimization on Google’s side
• You may be unknowingly excluded from certain optimizations

Reason 3: You may block more than you intend

Rules in robots.txt can have unintended consequences:

# Attempt to block parameter
Disallow: /*?gtm_latency=1

# BUT this rule may also block:
# /page/?gtm_latency=1&other_parameter=value
# /page/?important_parameter=something&gtm_latency=1

Depending on the implementation, you may accidentally block URLs with other, important parameters.

Reason 4: This treats symptoms, not the cause

Blocking in robots.txt is just an attempt to work around the problem, not solve it. Even if you block ?gtm_latency=1, what about other parameters that may appear in the future? ?utm_source=...? ?fbclid=...? ?gclid=...?

Proper canonical solves the problem for all parameters, once and for all.

What instead of robots.txt?

✅ Recommendation 3: Monitor server logs

Google Search Console is a great tool but doesn’t show everything. In the case of ?gtm_latency=1, I noticed that:

• URLs with the parameter appear in server logs
• But DO NOT appear in GSC reports

This means that if you rely only on GSC, you may miss the problem.

Why server logs are most important

Server logs are the only source of complete truth about what bots do on a site:

Log analysis tools

Summary

What started with one strange line in the server logs turned out to be a much bigger problem than I could have imagined 🙂

?gtm_latency=1 is an internal diagnostic parameter of Google Tag Manager that:

• Forces full telemetry of performance, tags, and Consent Mode
• Is added by the GoogleOther bot during crawling
• Theoretically shouldn’t end up in the main search index

Facts