Why has my website traffic dropped even though my Google ranking hasn't changed?

Who this applies to

This fits businesses whose organic traffic comes mostly from Google — especially expertise-led firms, consultants, agencies and B2B service providers whose buyers research before they click. It applies whether the drop was gradual or sudden, as long as the ranking positions themselves have held.

It doesn't apply if your rankings have slipped too, or if the drop lines up with a site migration, a broken analytics tag, or a seasonal dip you've seen before. Those are different problems with different fixes, and they're more common than this one. Rule them out first.

What's actually going on

Ranking is a snapshot. Your page holds a position for a query and largely stays there until the page or the algorithm changes. It's stable enough to put on a dashboard and check monthly.

Citation behaves nothing like that. Each time someone asks a question, the model retrieves candidates and selects a handful to build the answer from, and it does not reliably pick the same ones twice. A University of St Gallen study that asked the same prompts day after day found only about 34–42% of cited sources carried over between consecutive days. Roughly two-thirds of the sources changed overnight, with nothing changing on anybody's website.

The engines don't agree with each other either. In Kevin Indig's analysis of a 20,000-prompt sample, only 2.37% of cited URLs appeared across all three engines for the same prompt, while 91% appeared in just one. So “are we cited?” has a different answer per tool, and per day.

Meanwhile the answer itself absorbs the click. Someone who reads a sufficient answer at the top of the page has no reason to scroll to your result, however well it ranks. Your position didn't move. The thing sitting above it did.

Ranking and citation: what's the difference

They aren't unrelated. Google's AI experiences use Search's own index and ranking systems, so crawlability, indexing and page quality still do work in both places. But passing the ranking test doesn't carry you through the second one.

Where the loss concentrates: research, not buying

The most useful published breakdown comes from a University of Toronto study of 300 consumer-electronics queries, split evenly across informational, consideration and transactional intent. For consideration questions — the “who should I use for this?” stage — earned sources dominated, taking 59–86% of citations across the AI systems tested. Independent reviews and editorial coverage, not the brands' own pages.

For transactional questions, the same systems swung the other way: 52–68% of citations went to brand-owned pages. Once somebody knows what they want, your own site gets named again.

That's a consumer-electronics query set, so treat the numbers as direction rather than a rule for your sector. But the shape is worth knowing, because it gives you a diagnostic: a drop concentrated in research-stage traffic, with purchase-stage traffic holding up, points here. A drop spread evenly across both points somewhere else.

When this is the wrong explanation

• Direct or referral traffic dropped too. AI citation only touches the organic search channel, so a broad drop is a tracking or site problem.
• Your rankings did move, and nobody was watching the right keywords.
• The date lines up with a migration, a redirect change, a consent-banner change or a new analytics property.
• Your traffic is mostly transactional and branded — the end of the journey, where the Toronto data has AI systems favouring brand-owned pages anyway.

Common mistakes

• Reading a stable ranking as no change. The position held. What sat above it didn't.
• Checking AI visibility once. With only a third to a half of cited sources repeating day to day, a single check is closer to a coin toss than a measurement.
• Checking one AI tool and calling it done. Under 3% of cited URLs are shared across all three of the engines Indig sampled. You can look strong in one and be absent from the other two.
• Adding keywords in response. In the Princeton and IIT Delhi GEO study, keyword stuffing was the one method that made visibility worse. Adding statistics, quotations and cited sources were among the methods that helped, lifting visibility by up to 40% in their benchmark.

What to do about it

Start by finding out whether it's true for you, because most of the above is general evidence and none of it is about your business. Ask the questions your buyers actually ask, in the tools they actually use, and note whether you're named. Repeat it — a single sample tells you almost nothing.

Then work on the questions where you're missing, not on your whole site. What gets selected tends to be specific: a named person taking a clear position, a method described in enough detail to repeat, a number with a source attached, an answer phrased the way people actually ask. And where the gap is an editorial or review source rather than your own page, the useful move is a credible pitch, not another page on your site.

Frequently asked questions