AI Chatbot Traffic: What It Is, and How to Get More

Everyone’s talking about the traffic they’re losing to AI. Fewer people are talking about the traffic they’re gaining from it.

When ChatGPT, Perplexity, or Claude cites your content in a response, some of those users click through the citation and visit your website. That’s AI chatbot traffic, and it may convert better than most channels you’re probably already tracking.

Here’s which AI assistants send the most AI traffic, how to measure it, and how to get more of it.

chatgpt-vs-google.com, which tracks referral traffic across 74,752 websites, all AI chatbots combined sent 3.5 million visitors in March 2026—just 0.28% of total web traffic. Google sent 345.2M million, a 28.12% share.

A bar chart titled "Traffic Channels" comparing the percentage of web traffic share. Direct traffic leads at 40.10%, followed by Search (31.39%), Paid (17.17%), Social (7.34%), Unknown (5.73%), and AI Assistants at a small 0.28% share.

But Google’s share has been eroding. It dropped from 35.11% in June 2025 to 30.53% in March 2026—a 4.58 percentage point decline in ten months. At the same time, Detailed.com has reported that only 18% of large publishers experienced year-over-year traffic growth in early 2025, largely due to shifts in AI-driven search.

chatgpt-vs-google.com, tracking 74,752 sites.)

A dashboard titled "All Traffic Sources Comparison" for April 2026. A yellow box highlights the "AI Assistants" column, showing traffic for ChatGPT (2.4M), Meta AI (521.1K), Perplexity (200.9K), Gemini (223.5K), and Claude (97.4K).

But the growth numbers tell an interesting story, too. Claude grew 153.5% month-over-month in March—attributed to Anthropic expanding its web search capabilities—with an average growth rate of 30.7% over the tracked period.

Gemini grew 21.2% in March, averaging 12.8% growth. ChatGPT is the volume leader, but it’s the slowest-growing of the four, with a 1.4% average growth rate. Perplexity sits at 2.9%.

A bar chart titled "Month-over-Month Growth Comparison" showing percentage changes in visitors. A tooltip for March 2026 highlights massive growth for Claude (+154.7%) and Gemini (+21.9%), while ChatGPT (+0.4%) and Meta (-14.5%) show much lower or negative growth.

0.5% of total visitors but drove 12.1% of signups—a 23x higher conversion rate than organic search.

We also track sign-ups self-attributed to AI at onboarding: March 2026 was a record month, with ChatGPT credited by 1978 new users, Claude by 2,836, Gemini by 619, and Perplexity by 67—with Claude growing fastest.

I’ve seen many stories like that shared directly by companies on LinkedIn. Here’s another one from Simon Heaton, Director of Growth Marketing at Buffer: conversion rates 185% higher than organic search.

A screenshot of a LinkedIn post by Simon Heaton showing a bar chart titled "Aggregate Conversion Rate Comparison." It shows that LLM-driven traffic has a 20.15% conversion rate compared to 7.06% for Organic Search, representing a 185.4% relative uplift.

One of the reasons why this type of traffic often converts better is that the AI already does some product research for the customer. Someone who clicks a citation in a chatbot response has already received an AI-generated answer and decided they want more. That’s a deliberate click. Compare it to someone who followed a display ad or clicked a social post while scrolling.

In fact, Kevin Indig’s study of 48 participants making high-stakes purchases found that 64% of AI Mode users clicked nothing at all—they got their answer without leaving the chatbot. Of the 23% who did click through, most visited to confirm a choice already made, not to explore options. So, AI-referred visitors who land on your site have typically already decided. It’s more verifying before the purchase than browsing.

Ahrefs Web Analytics has AI traffic reporting built in—no custom segments, no consent banners. Add the JavaScript snippet to your site, and here’s what you can do (full guide):

Here are a few ideas on exploring that data in a practical way.

See your AI traffic vs. other channels. This shows AI as a share of total sessions alongside organic, direct, and referral—useful for deciding how much strategic weight to give the channel. You can also compare behavior metrics (bounce rate, session duration) directly against other channels here.

Go to the Overview report
Under “Traffic Sources,” select the Channels tab
Click View More to open the full channel report
Apply the LLM channel filter to isolate AI traffic

Ahrefs Web Analytics "Overview" dashboard. The "Traffic sources" section shows a list of channels with an orange arrow pointing to a "View more" link under the "Channels" tab to isolate specific traffic data.

Find which AI platforms send you the most traffic. You’ll see a breakdown by platform (ChatGPT, Gemini, Perplexity, Copilot, Claude) with trend lines over time. Switch to Relative view to spot share shifts — useful for catching when one platform starts outpacing another.

Keep the LLM filter active
Under “Traffic Sources,” switch to the Sources tab
Click View More

Ahrefs Web Analytics "Sources" report with a filter applied for "Channel: Is LLM." It shows a line graph and table comparing unique visitors from ChatGPT (2.6K), Gemini (881), Perplexity (165), Copilot (8), and Claude (1).

Find which content gets cited most. The Pages report shows your most-visited content from AI with bounce rate and time-on-page. Low bounce + high time on page signals content that AI visitors actually engage with after clicking — worth doubling down on.

Keep the LLM filter active
Under “Pages,” select Top Pages → View More

Ahrefs Web Analytics "Overview" dashboard highlighting the "Pages" section. An orange arrow points to the "View more" link under "Entry pages" to find which specific content is being cited most by AI.

Ahrefs Bot Analytics

If you also want to monitor AI crawler activity—not human referrals, but the bots actually crawling your pages—Ahrefs Bot Analytics tracks which bots visit your site, how often, and which pages they crawl, broken down into 12 categories, including a dedicated AI bots filter.

Ahrefs "Bots" report with the "AI bots" filter active. It displays a line graph for various crawlers including ChatGPT-User (1.1M), Meta-ExternalAgent (174K), and ClaudeBot (102K), showing how frequently AI systems crawl the site.

This matters because crawl access is a prerequisite for citation. If GPTBot or ClaudeBot isn’t crawling your content, those platforms can’t cite it—which means no referral traffic downstream. Bot Analytics shows you whether they’re getting in, how frequently, and which pages they’re hitting.

Bing Webmaster Tools

If Copilot is on your radar, Bing Webmaster Tools has a dedicated report for AI-generated response clicks. Under the Search Performance section, you can filter by “Generative” to isolate traffic that came from Copilot’s AI answers—broken down by page and by the prompt that triggered the response.

It’s one of the few places you can see the actual user queries that led to a click, which makes it useful for understanding what topics your content is being surfaced for—and what you’re missing.

Additionally, you can use this data as an AI visibility proxy for Google Search. There’s a high chance that people will find you in other AI systems through the same prompts as you see in Bing Webmaster Tools.

Branded web mentions are the factor that correlates most strongly with brand appearance in AI answers (Spearman correlation: 0.664). More mentions across the web mean more training signal for the LLM, which means a stronger association with your topic.

To get more meaningful online mentions, you should focus on:

PR and media coverage. Arguably more valuable in the AI era than it was before. What authoritative outlets say about you gets ingested by AI systems and shapes the answers they generate, so coverage becomes part of the training and retrieval substrate. The stronger the consensus across publications, the more consistently you’ll show up in AI responses, because the model gets more confident about the narrative around your brand.
Industry “best of” lists. They aren’t always the highest-quality content on the web, but AI systems still lean on them heavily for product recommendations because rankings and reviews look like the most directly relevant piece of information for a buying question. Last time we checked across all source links AI systems pull from, “best X” blog lists make up 43.8% of all page types. Getting onto the right roundups—the ones real publications maintain, not the SEO-bait ones—is one of the most direct paths to AI citations.
Reddit and YouTube. Reddit and YouTube have a special place in AI’s pipeline — they sit on a dedicated retrieval lane (separate API feeds layered on top of regular search), pulled in at huge volume so the model can sense “what real people say.” They almost always show up in the top-cited domains because of that volume.

A straightforward way to start with this is to find your AI mention gaps using Ahrefs Brand Radar (AI answers that mention your competitor but not you).

Enter your brand and add competitors (or use the AI suggest feature)
In the Overview tab, compare mentions, impressions, and AI Share of Voice per platform
Click into any platform, select Others only to filter to queries where competitors are mentioned but you aren’t
Open Cited Domains to see which outlets are driving competitor mentions—those are the publications to target with PR

Build content AI wants to cite

Four things drive whether a page earns citations once crawlers can reach it: format, structure, topical coverage, and freshness.

Pick formats AI leans on. Across sites, “best” roundups (7.06% of AI traffic), how-to guides (6.35%), “top” lists (5.5%), “vs” comparisons (4.88%), and product/service pages (4.5–6.8%) consistently outperform.

The pattern across our own three months of AI traffic data agrees: how-to guides and original data studies pull the most views and the deepest engagement, while definitions and product pages do quieter but valuable work for navigational and bottom-funnel queries.

The takeaway isn’t to pick one format—it’s to cover several so you catch different user intents.

Cover the long tail. Studies of real ChatGPT conversations found the average prompt is 42 words long, and 75% are commands rather than questions (“how to create,” “best way to track”). Content clusters that cover every angle of a topic—not just the head term—match how people actually prompt AI.

Two ways to find the gaps in Ahrefs:

Keywords Explorer, enter your main topic, go to Matching Terms, group by Clusters by Parent Topic, open the Questions tab, and filter by your domain to see which angles you cover and which you don’t.

Ahrefs "Keywords Explorer" showing a treemap of "Clusters by Parent Topic" for the topic "coffee." It highlights the "mushroom coffee" cluster and points to a "Position" column to show which specific questions the user's site does not currently rank for.

Command-based gaps: in Competitive Analysis, compare your domain against two or three competitors and filter for keywords containing “how to,” “create,” “track,” or “make.” Those are the task-completion queries competitors may already own.

Ahrefs "Content gap" report comparing one domain against three competitors. An orange box highlights keywords containing task-based phrases like "how to," "create," or "track" that competitors rank for but the target site does not.

Lead with the answer. AI systems only consider the first 30 passages of a page for embeddings, and passages get retrieved individually, so each section needs to make sense on its own. Don’t bury the main point. Answer first, then expand.

Keep content fresh. AI assistants cite content that is 25.7% fresher than what appears in organic search, with a 13.1% preference for recently updated pages. HubSpot updated one post on small business ideas and earned 1,135 new AI Overview mentions from that single refresh. Regular updates aren’t just SEO hygiene—they’re a direct citation signal.

Agent A, Ahrefs’ built-in AI assistant, handles it for you.

It connects directly to the same data sources mentioned throughout this article, including Keywords Explorer, Brand Radar, Web Analytics, and Bot Analytics. From a single prompt, it can run multi-step workflows that would normally take a full research session.

A screenshot of Ahrefs' "Agent A" AI assistant interface. A user has entered the prompt: "Which pages haven’t been updated in the last 6+ months but still receive AI traffic?" and the agent is processing the request for the domain ahrefs.com.

Here are a few prompts that work especially well for an AI citation strategy:

“Compare our ‘how-to’ content against [competitor] and [competitor]. Show the highest-volume keyword gaps.”

“Find every page on our site ranking for AI-cited queries with a bounce rate above 70%. Those are good candidates for an answer-first rewrite.”

“Which pages haven’t been updated in the last 6+ months but still receive AI traffic?”

“Using Brand Radar, show domains where competitors are being cited but we aren’t, ranked by frequency. Those are potential PR opportunities.”

Final thoughts

Most “AI SEO” advice right now is people pattern-matching on a channel that barely exists yet. 0.28% of total web traffic is not a strategic priority for most businesses. But it’s a strategic priority for the ones who notice the conversion math: a channel converting, in some cases, 23x better than organic search, is worth ten times the effort its raw volume suggests.

Don’t rebuild your content strategy around it yet. Do set up tracking, ship the formats AI leans on, and check your analytics. That’s enough to be ready when the volume arrives.