# I Lost 93% of My Clicks. AI Cites Me More Than Ever.

**Last updated:** 2026-06-17  
**Source:** https://www.lawrencehitches.com/clicks-to-citations/

---

One page on my site went from **1,236 clicks a month to 82** in six months. Same page. No de-indexing, no penalty, no rewrite gone wrong. Google's AI Overviews simply started answering the question before anyone needed to click. The three-channel model this page describes is now tracked on the [AI SEO studies](https://www.lawrencehitches.com/ai-seo-studies/) page with its monthly reads.

Clicks are collapsing while AI citations rise: the click-to-citation shift in first-party data. Download PNG

At the same time, AI crawlers became **26% of my entire traffic**, and they cite my content 46 times for every single visitor they send back.

That is the shift, in two numbers from my own data: **clicks are collapsing, and citations are becoming the currency.** Here is exactly what happened, with the first-party numbers, and what I think you should do about it.

## The page that died (in slow motion)

My most-trafficked page was [utm_source=chatgpt.com explained](https://www.lawrencehitches.com/utm-source-chatgpt-explained/). For most of 2025 it carried the site. Then it fell off a cliff:

Decay chart Monthly clicks, one page, Nov 2025 to May 2026 Nov 1,236 Dec 424 Jan 149 Feb 166 Mar 102 Apr 54 May 82

A 93% drop in monthly clicks, with the page still indexed and still ranking.

The query behind it, "what does utm_source=chatgpt.com mean", is exactly the kind of question an AI Overview answers in one sentence. So Google does. Microsoft Clarity shows people who do land on the page spend 79 seconds on it but scroll only 19.5% down. They read the one-line answer at the top and leave. They never needed the other 2,000 words.

Two things drove the fall: AI Overviews swallowing the informational query, and the novelty wearing off (when the parameter first appeared, everyone Googled it; now it is normal, so the search demand itself is shrinking). Neither is fixable with more content.

## This is not just my page

I checked my decay against the wider data, and it lines up. AI Overviews now appear on more than 25% of Google searches. Independent 2026 studies put the click-through drop on those queries between 34% and 61%, and the zero-click rate where an AI Overview appears at 80% or higher. Publishers are reporting Google referral traffic down around 38% year on year.

So the macro picture matches the micro one on my own site. The era where a strong informational ranking reliably produced clicks is ending for the kind of query a machine can answer in a sentence.

## Meanwhile, the bots moved in

Here is the part most traffic reports miss, because they filter bots out. Over one week, AI crawlers and assistants made up **26.3% of all requests to my site**: 13,565 AI bot hits. The breakdown:

Bot breakdown AI bot requests by operator (one week) OpenAI 61.2% Meta 15.9% Huawei 10.4% Amazon 5.9% Apple 2.3% Anthropic 1.7%

61% of the bot activity is AI assistants answering live queries; 35% is crawlers harvesting for training.

Read that again. More than a quarter of my traffic is now machines reading my content to answer other people's questions, mostly inside ChatGPT. My human Google clicks fell while machine reads climbed. The audience did not leave. It changed shape.

## They take a lot and give little back. Mine gives back more than most.

The fair criticism of AI crawlers is that they take content and return almost nothing. The industry measures this as a crawl-to-referral ratio: pages scraped per visitor sent back. The published numbers are brutal.

Ratio comparison

| Crawler | Scrapes per referral (industry) |
|---|---|
| Google (Search) | ~5 : 1 |
| PerplexityBot | ~195 : 1 |
| GPTBot (OpenAI) | ~1,091 : 1 |
| ClaudeBot (Anthropic) | ~23,951 : 1 |
| My site (overall) | 46 : 1 |
| My site (OpenAI) | 7 : 1 |

My overall ratio is 46 to 1. For OpenAI specifically it is 7 to 1, against an industry GPTBot average of roughly 1,091 to 1. That is the most interesting number I have. The absolute referral count is still small, so I am not claiming a flood of traffic. But the efficiency is an order of magnitude better than the web average, and it is consistent: content built to be cited gets cited, and a meaningful share of those citations send a real person back.

In other words, the crawlers are not the enemy. The badly-structured site that gets harvested and never cited is the loser. The site that gets cited and occasionally referred is winning a game most people are not even scoring.

## So is AI traffic worthless? No. The payoff moved.

If you measure AI search only in clicks, my year looks like a decline. Measure it in citations and it looks like the opposite.

On Bing's new AI visibility data, my overall citation share of authority is **12.72%** of all citations for the queries I appear in, and on my core commercial intents (enterprise SEO platform comparisons, WooCommerce SEO evaluations, Copilot ranking), I hold **40% to 90%** share. When someone asks an AI which enterprise SEO platform to choose, my analysis is often the source it leans on.

Does that convert? Here is the proof I keep coming back to. An enterprise conference booked me as a speaker, over 400 attendees, through my website inquiry form. I asked how they found me. They had asked Microsoft Copilot about AI search experts, and my name came up. That booking did not show up as a click in any analytics tool. It came from a citation.

That is the whole thesis. **Clicks were how search paid you. Citations are how AI search pays you**, in visibility, in brand recall, and in the occasional high-value lead that no dashboard will ever attribute.

## What the click-to-citation shift means for you

This is not a reason to panic, and it is definitely not a reason to block AI crawlers. It is a reason to change what you measure and what you build.

- **Stop scoring AI search in clicks alone.** Track citations and citation share, branded search volume, and direct traffic alongside referrals. The win is mostly in the dark funnel now.
- **Measure your citation share.** Bing Webmaster Tools added free AI Visibility Insights in June 2026, citation share, intents, and topics, for the Bing and Copilot index that ChatGPT Search also uses. Start there. I cover the full setup in [how to track AI search rankings](https://www.lawrencehitches.com/ai-rank-tracking-tools/) .
- **Build for extraction.** The content that earns citations leads with the answer, uses clear headings, names specifics, and includes data. The same structure that wins a citation also wins a featured snippet. You are not writing for the machine, you are writing clearly, and the machine rewards it.
- **Do not block the crawlers.** Blocking GPTBot or ClaudeBot to "protect" your content removes you from the answers your buyers are reading. That is the opposite of what you want.
- **Accept that informational pages will bleed clicks.** If a query can be answered in a sentence, an AI Overview will answer it. Put your effort into content with a point of view, first-hand data, or a decision the reader has to make, the stuff a one-line answer cannot replace.

My biggest page is never coming back to 1,236 clicks a month. I have made peace with that. The same content quality that made it rank is now making me the cited source across a category, and that is turning into work. Different currency. Same fundamentals.

If you want to see where your own content stands, run a page through the free [low quality content checker](https://www.lawrencehitches.com/low-quality-content-checker/), then check your citation share in Bing. The first tells you if you deserve to be cited. The second tells you if you are.

The same shift drives [query fan-out](https://www.lawrencehitches.com/query-fan-out-seo/): AI grounds in whoever answers the most sub-queries, and rewards citations over clicks.

AI Search Metrics by Confidence Level What each metric tells you and how to measure it, sorted by how far you can trust the number.

| Confidence | Metric | What it tells you | How to measure |
|---|---|---|---|
| Observed | AI-referred sessions | Whether your known AI traffic is growing or shrinking. The floor, not the ceiling. | Sessions from AI referrers (chatgpt.com, claude.ai, perplexity.ai) in a dedicated GA4 channel group. |
| Observed | AI conversion rate | Whether AI traffic converts as well as organic. A quality check, not just a volume one. | AI conversions divided by AI-referred sessions, compared against organic. |
| Observed | Bot-grounding share | How much of your traffic is AI crawlers ingesting your pages. The input to citation. | Verified AI-bot requests divided by total requests, from your CDN or server logs. |
| Observed | Crawler-to-Assistant ratio | Whether a page has converted from indexed to cited. The grounding funnel. | AI Assistant hits divided by AI Crawler hits per URL. A rising ratio means it is being cited live. |
| Proxy | Bing AI citation share | Your citation visibility across Microsoft and Copilot surfaces. | Bing Webmaster Tools AI performance: your share of authority on cited queries. |
| Proxy | Branded and direct lift | Hidden AI demand that referral data alone misses. | Branded clicks and direct sessions versus baseline, on pages AI surfaces often. |
| Proxy | AI discovery rate | AI influence on users who arrive branded or direct and would otherwise be invisible. | Survey question "did you first hear about us via an AI assistant?" divided by responses. |
| Modelled | AI-influenced pipeline | A planning estimate, not attributed proof. Always report it as a range. | Influenced visits above baseline times your historical conversion rate. |

The closer a metric sits to Observed, the more you can trust it. Modelled numbers are for planning, not proof. Measure grounding, not just clicks.

Skip the re-prompting with the [Claude skills for SEO](https://www.lawrencehitches.com/claude-seo-skills/), free and open source.

## Frequently asked questions

### Do AI Overviews reduce traffic?

Yes, for queries an AI Overview can answer directly. Independent 2026 studies put the click-through drop between 34% and 61% on affected queries, with zero-click rates above 80%. On my own site, one informational page lost 93% of its monthly clicks over six months while still ranking. Transactional and decision-stage queries are far less affected.

### How much of my traffic is AI bots?

On my site, AI crawlers and assistants are 26% of all requests, led by OpenAI at 61% of that bot activity. Industry estimates put AI crawlers at roughly 4 to 8% of requests across the web and rising fast, but content-heavy SEO sites like mine see much higher shares.

### What is a good crawl-to-referral ratio?

Lower is better, it means the crawler sends back more visitors per page it takes. Google sits near 5:1 because Search exists to send clicks. Pure AI crawlers are far worse: GPTBot around 1,091:1 and ClaudeBot around 23,951:1 by industry estimates. My site runs 46:1 overall and 7:1 for OpenAI, which suggests citation-friendly content gets referred back far more efficiently than average.

### Are AI citations worth anything if they do not send clicks?

Yes. A citation puts your brand in front of someone at the moment they are making a decision, even if they never click. It drives brand recall, branded search, and the occasional high-value lead. I booked a 400-person speaking engagement from a single Copilot citation that never registered as a click.

### How do I track AI citations?

Use Bing Webmaster Tools' AI Visibility Insights (free, launched June 2026) for citation share, intents, and topics on the Bing and Copilot index. Pair it with GA4's AI Assistant channel for referral traffic and Search Console's generative AI report for impressions. Full walkthrough in [how to track AI search rankings](https://www.lawrencehitches.com/ai-rank-tracking-tools/).

---

*Lawrence Hitches is an AI SEO consultant in Melbourne and Chief of Staff at StudioHawk, Australia's most awarded SEO agency. Agent edition of this site: https://www.lawrencehitches.com/md/index.md*