Traffic dropped but rankings stayed the same: 6 causes

This is now the single most common question in SEO reporting, and it is a genuinely confusing one, because every instinct says rank and traffic move together. They used to. A position number describes where you sit in the list of organic results, and it says nothing at all about how much of the screen that list occupies, how many people scroll far enough to see it, or whether their question was already answered before they got there. When any of those change, clicks move and rank does not.

1. An AI Overview appeared above your result

Start here, because the numbers are large and the effect is invisible in a position report. Ahrefs data collected in December 2025 found the top organic result loses roughly 58% of its click-through rate when an AI Overview is present, compared with 34.5% measured in April 2025. BrightEdge tracked AI Overviews on 48% of monitored queries in February 2026, up 58% year over year. Your position stays at 3. The page it sits on now answers the question before anyone reaches you.

The tell is a step change rather than a slope. Clicks on one page or one keyword group fall over a few days and then flatten at a new, lower level, while impressions hold up or even rise. If you have a dated feature history you can match the step to the day the block appeared. If you do not, this is exactly the gap that AI Overview tracking closes: a flag on every daily check recording whether the feature was there, stored next to your position so the two lines can be read together.

One important nuance before you panic. The loss is not uniform. Informational queries take the worst of it, because a definition or a how-many question can genuinely be finished in the answer block. Commercial and transactional queries lose considerably less, since somebody comparing vendors or about to spend money still clicks through to check pricing, reviews and specifics. Sort your affected keywords by intent before you rewrite anything.

2. Someone else won a featured snippet or a new SERP feature landed

AI Overviews get the headlines, but the older feature blocks still move plenty of clicks. A competitor winning the featured snippet above you takes the top of the page without changing a single organic position. A people-also-ask box expanding after result two pushes everything below it down a full screen on mobile. A video carousel, a shopping row or a local pack does the same. None of it registers as a rank change.

That table is most of the diagnosis. What makes it work is having the SERP stored on the dates in question, which is the difference between SERP tracking and plain position checking.

3. Google rewrote your title or description

Google rewrites a large share of title links, and it also regenerates description snippets per query. A rewrite that drops your strongest phrase, or a snippet that now answers the question in the SERP itself, lowers click-through rate at an unchanged position. It is easy to miss because nothing in your CMS changed. Search your target keyword, compare the snippet Google is showing against the title tag you wrote, and if they differ, look at whether the version being shown is doing your page any favors.

4. Demand for the query fell

Fewer people are searching, so fewer people click, and your position is doing its job perfectly. This looks nothing like the AI Overview pattern: impressions fall alongside clicks, the decline is gradual, and it usually covers a topic cluster rather than a single page. Check Search Console impressions and a year over year view before anything else, because a genuine seasonal trough gets misdiagnosed as an algorithm penalty every single quarter, and teams burn weeks rewriting pages that were never broken.

5. Your analytics broke, not your traffic

A surprising share of dramatic traffic drops are measurement failures: a tag removed during a deploy, a consent banner change, a filter or a bot rule that started excluding real sessions, a property switched. The signature is a cliff on one exact date with no corresponding move in Search Console clicks. Always reconcile the two sources before you accept the drop is real, and if your reporting flows through a warehouse before it reaches the dashboard people look at, it is worth tracing where those numbers actually come from rather than trusting the final chart. Ruling this out takes twenty minutes and it saves the occasional embarrassing meeting.

6. You are averaging away a real ranking loss

Sometimes the rankings did move and the report hid it. Search Console average position is averaged across every query and every location that produced an impression, so losing position 3 on your best commercial term while gaining ground on fifty long-tail variants can leave the average almost unchanged. Look at your revenue keywords individually rather than at a portfolio average, and at the specific pages that lost clicks. We covered this trap in detail in Search Console average position vs a rank tracker.

Why is my traffic going down but my rankings are the same?

Because rank measures your place in the organic list, and traffic depends on how much attention that list receives. An AI Overview, a featured snippet you do not own, a shopping row or an expanded people-also-ask box all take attention from above without taking your position. Confirm by looking at the current results page for the keyword and comparing it against what the page looked like before the drop.

Why are my impressions up but my clicks down?

That combination points strongly at a click-through rate problem rather than a ranking problem, and the usual culprit is a feature block absorbing the answer. You are being shown to the same people or more, and fewer of them are choosing your result. The other candidates are a rewritten title link, a snippet that gives everything away, or newly ranking for broader terms where you sit lower down the page.

Can you recover traffic after an AI Overview appears?

Partly, and rarely all of it. Being cited inside the answer block wins back some clicks, and the observed pattern is that generated answers lean on pages already ranking in the top 10, so conventional ranking work remains the entry ticket. Beyond that, structure helps: answer the exact question in the first 40 to 60 words under a heading phrased the way people search it, keep facts current and dated, and use tables for comparisons. The more reliable strategy is reallocation. Move effort toward commercial queries that still send clicks, and stop investing in informational terms that are now answered in place.

The diagnostic order that saves the most time

Work from cheapest to most expensive. Look at the live SERP for the affected keyword. Compare Search Console clicks against your analytics to prove the drop is real. Check whether impressions moved with clicks or against them. Compare year over year. Only then start auditing the page itself. Most teams do this in reverse, spend a week rewriting content, and discover afterwards that an answer block landed in week one.

The structural fix is to stop finding out late. A daily record of position plus features per keyword means the next time this happens you get an alert on the day the page changed, not a question from a client six weeks later. Set thresholds on rank drop alerts for position moves, and watch the feature column for the drops that never show up as a position move at all. If several clients or brands are involved, the ranking reports export turns the same history into the explanation you send them.