Best URL Shorteners That Filter Bot Clicks From Analytics

A short link reports 400 clicks. That number is a blend of at least four populations that behave nothing alike, and the useful question is not how big it is but what it is made of.

Real people

The population you actually want. Everything else is noise dressed as signal, and a safety checker hitting your link counts as noise too.

Crawlers and scanners

Search engine bots, security scanners checking whether your link is malicious, uptime monitors, and scrapers. These arrive fast, often in bursts, and can outnumber humans entirely on a link that has been posted somewhere public.

Link previews

Slack, WhatsApp, iMessage, Discord and Twitter all fetch a URL to build a preview card the moment someone pastes it. One person sharing your link in a busy channel can generate dozens of fetches before a single human clicks.

AI assistants

The newest category and the one nobody had to think about two years ago. An assistant fetching your page to answer somebody's question is not a visitor, but it is also not a crawler indexing you. It represents a person who asked about your thing, which makes it genuinely interesting rather than merely noise.

We checked nine shorteners against their own documentation in August 2026. All of them filter. The interesting split is disclosure.

We were wrong about this being rare

Worth stating outright, because we had it in our own notes as a differentiator: showing the bot count is common. Linkly puts it directly beside the total. TinyURL treats bot and human as two labelled metrics rather than filtering at all. Short.io surfaces it in three places. If a comparison tells you only one vendor does this, it has not checked.

TinyURL's approach is different in kind

TinyURL labels rather than removes: it captures Bot Clicks and Human Clicks as separate metrics and exposes a per-click flag through its API. That is arguably more honest than filtering, since nothing is discarded, and it does mean the headline total on a TinyURL link includes bots unless you read the split.

A filtered number you are never shown is just a smaller number you still cannot verify.

If bot disclosure is common, what is not? Treating AI-assistant traffic as its own thing.

Why it is worth separating

An assistant fetching your link is a signal about demand, not just noise. Somebody asked a question, an AI went to your page to answer it, and no human ever saw your analytics-tracked visit. Lumped into "direct traffic" that is invisible. Filtered out entirely, it is deleted. Named as its own channel, it is a measurement of how often you are the answer.

What Flyn does

The bot-excluded human count is the headline number on every plan including free, and the traffic-quality view showing human beside filtered-bot is free too. Pro adds the AI Assistants channel alongside the geo, city, device, OS, browser and referrer breakdowns, so assistant traffic sits in the same report as everything else rather than being thrown away.

What the rest do

Rebrandly, Short.io, Linkly, Cuttly, TinyURL, Dub and T.LY do not document AI-assistant traffic as a category anywhere. That is not an accusation of a missing feature; it means the vendor does not describe it, so do not buy expecting it.

Bitly deserves its own section because the easy summary is wrong, and we published the easy summary ourselves before checking properly.

In the product, it is gone

Bitly's help centre is direct: agentic traffic is filtered out of your click data the same way crawlers are, and the numbers on the Links page and Analytics dashboards reflect visits from people.

In the API, it is a number

Bitly also added an agentic_traffic dataset to its group analytics, described in its own tools reference as counting clicks from AI agents and assistants, always returning a fixed 90-day total. So the data exists: group-level, one rolled-up figure, reachable through the API or its MCP server.

What you can and cannot say about it

You cannot say Bitly discards AI traffic, because it does not. You can say you will not find it per link, over time, or anywhere in the dashboard. We corrected our own article on which clicks came from ChatGPT for exactly this, and the correction is worth more than the original claim was.

The reason any of this matters is that unfiltered numbers change what you do, usually in the direction of wasting money.

A/B tests break first

Bots do not distribute evenly across a weighted rotation. A scraper that hammers one branch will hand you a winner that is not a winner. If your rotator reports per destination but your click count includes crawlers, the comparison is between one real page and one crawled page.

Affiliate campaigns look better than they are

Affiliate links attract scrapers, and a click count inflated by them makes a poor offer look promising. This is why cloaked affiliate links deserve a filtered count underneath them.

Paid campaigns get scaled on fiction

If you judge a channel by clicks and a share of those clicks are automated, you will scale spend into a number that cannot convert. Our guide to click fraud prevention covers the diagnostics for spotting it.

If you just want the number, free

Flyn, Linkly and TinyURL all show it without a paid plan, and Short.io surfaces it in three places. Any of these is a reasonable answer, and we are not going to pretend ours is the only one.

If you need to audit a suspicious spike

Cuttly and Linkly break filtered traffic down by bot name, which the others do not. If you are trying to work out whether one scraper is responsible, that breakdown is the feature you want, and neither is us.

If AI traffic is the question

This is the one place we would argue Flyn is genuinely ahead: a named channel, per link, sitting in the same report as your other breakdowns. It is a Pro feature at $12/month, or $9 billed annually. You can start on the free plan and see the human-versus-bot split before deciding.

If you are on a silent filterer

Bitly, Dub and T.LY all filter without showing you the number. That is a defensible engineering choice and it does mean you have no way to sanity-check the total. If your traffic looks strange, you will have nothing to check it against.

Once you can see the split, a few habits make it useful rather than decorative.

Watch the ratio, not the count

Bot volume rises with exposure, so a growing bot count on a link that is being shared more is normal. A sudden change in the share is the signal. A link that was 15% bots and is suddenly 60% has been found by something.

Expect previews to spike at publication

The first minutes after you post a link are dominated by preview fetchers. Judging a campaign on its first hour of raw clicks will always flatter it.

Do not chase zero

No filter reaches zero, and a vendor claiming it does is overselling. The goal is a number stable enough to compare week over week, not a perfectly clean one.

Keep the human number as your reported number

Whatever you put in a report to a client or a manager should be the filtered human count, with the bot figure available if anyone asks. That is the whole reason disclosure matters more than filtering: you can defend a number you can decompose. Our guide to tracking clicks covers the rest of the reporting surface, and the clicks API returns the same split programmatically.