Checked 30 October 2026
Gartner’s forecast that traditional search engine volume would drop 25% by 2026 can’t be marked right or wrong, because it never said what it was measuring. On the closest public evidence, Google’s own statements about query growth, it looks like a miss. On publisher traffic, which Gartner didn’t forecast, it looks closer to right.
It’s still one of the most repeated numbers in AI search. It appeared on six of the 20 roundups in my AI search statistics audit. 2026 has arrived, so it’s time to check it with the same rules I use for my own calls.
What Gartner actually said
The press release is dated 19 February 2024 and runs to about 320 words. Its opening line says traditional search engine volume will drop 25% by 2026, with search marketing losing share to AI chatbots and virtual agents. Alan Antin, a Gartner VP analyst, is quoted saying generative AI is becoming a substitute answer engine for queries that used to go to search engines.
Read against this site’s rules for a checkable claim:
| What a checkable call needs | Gartner’s release |
|---|---|
| What’s measured | “Traditional search engine volume”, undefined |
| Baseline | Not stated |
| Geography | Not stated |
| Who measures it, and how | Not stated |
| Check date | “By 2026” |
The word “volume” appears once. “Traffic” doesn’t appear at all. The full research behind the release is for Gartner clients only, so some of this may be defined there. Publicly, it isn’t.
The evidence against
Google is the only company that knows its own query volume, and in each of its 2026 earnings rounds it has described query growth, not a fall.
| When | What Google said | Source |
|---|---|---|
| March 2025 | More than 5 trillion searches a year | Search Engine Roundtable |
| Q2 2026 | Search and other revenue up 17% year on year; AI features driving query growth | Search Engine Journal |
| Q3 2026 (28 Oct) | Sundar Pichai: year-on-year query growth is the highest in five years | Search Engine Roundtable |
These are self-reported, global and unquantified. Google gives a direction and no figure, and doesn’t separate AI Mode queries from classic ones. If AI Mode counts as “traditional search”, Gartner’s distinction collapses. If it doesn’t, Google hasn’t published the split.
SparkToro, using Datos’ US desktop panel, found searches per searcher on Google grew 21.64% in 2024 against 2023. That’s per user, desktop only, and SparkToro now flags the post as possibly out of date.
The evidence for
| Measure | Finding | Why it isn’t Gartner’s number |
|---|---|---|
| Apple’s Eddy Cue, May 2025 trial testimony | Searches in Safari fell for the first time in 22 years in April 2025 (9to5Mac) | One browser, one month, no size given. Google said total queries were still growing |
| Datos/SparkToro, Q4 2025 | US desktop searches per Google user reported down about 15–20% year on year | Per user, desktop only, US only. The two write-ups I read disagree on the exact figure; the report itself is gated |
| Fractl, June 2026 | 29% of search volume across 1,010,848 keywords sits in declining keywords (Fractl) | See below |
Fractl is the only study I found that tests Gartner directly, and it’s the one most likely to be quoted as proof. Its headline says it found “a 29% drop”. Its own text says something different: 29% of search volume sits in keywords that are declining, with a keyword counted as declining if it lost more than 15%. That’s the share of volume in shrinking keywords, not a 29% fall in total volume. It also uses Semrush’s modelled keyword volumes for US keywords with at least 10,000 monthly searches.
The traffic reading
Some coverage at the time, including MediaPost, reported the forecast as covering traffic to publisher sites as well as search volume. The release doesn’t say that. But if that’s what people heard, there’s more support for it: SparkToro reports that 68.01% of US Google searches ended without a click in January to April 2026 (SparkToro), against 58.5% in its 2024 study, which used a different panel. Searches can rise while clicks to websites fall. That’s the smaller click this site has been tracking.
The verdict
| Reading | Verdict | Confidence |
|---|---|---|
| As written (traditional search engine volume, total) | Can’t be scored | Gartner gave no definition, baseline or source |
| Closest public measure (Google’s query growth) | Looks like a miss | Low: Google reports direction, not numbers |
| Searches per user, US desktop | Partly right in direction, short of 25% | Low: one panel, one device type |
| Clicks to websites (not what Gartner forecast) | Right in direction | Medium: several independent panels agree |
If this were my call, I’d mark it void on the board and leave it visible. That’s what happens when a forecast skips the line that says what would count as being right.
The useful lesson isn’t whether Gartner was right. It’s that a number with no definition can be quoted as correct by anyone, for years, whatever happens.
What a scorable version looks like
Here’s the same idea written so it can be marked:
By 31 December 2026, Google’s reported annual search count will be lower than its last reported annual figure (more than 5 trillion, stated March 2025). Right if Google publishes a lower annual figure by then. Wrong if it publishes a higher one. Void if it publishes none.
It’s narrower, and less quotable. It can also be wrong, which is the point.
Method
The Gartner release was opened in a browser on 30 October 2026 and read in full. Every other source was opened the same day. Google statements come from earnings coverage that quotes the call; I didn’t use the earnings transcript directly. Where two write-ups of the same gated report disagreed (Datos Q4 2025), I’ve given the range and said so. Fractl’s figures were read on Fractl’s own page.
Limits
- No public dataset measures total search engine volume across engines and devices. Every row above is a proxy.
- Google has an interest in saying search is growing. It’s also the only party with the data.
- I haven’t read Gartner’s client research. If it defines the measure, this verdict could change.
- “Looks like a miss” is a judgement on proxies, not a score.
Related: How to write a claim you can mark wrong · AI search statistics, traced to the source · The future of search is a smaller click · I said I’d check my World Cup search predictions in August. I can’t.
Sources
Opened on 30 October 2026.
- Gartner: Gartner Predicts Search Engine Volume Will Drop 25% by 2026 (19 February 2024)
- MediaPost: Traditional search forecast to fall 25% by 2026
- Search Engine Roundtable: Alphabet Q3 2026 earnings (28 October 2026)
- Search Engine Journal: Google search revenue growth eases after a year of acceleration (Q2 2026)
- Search Engine Roundtable: Google 5 trillion searches per year (March 2025)
- SparkToro: Google search grew 20%+ in 2024
- SparkToro: In 2026, less than one third of Google searches still send a click
- 9to5Mac: Eddy Cue on Safari search decline (8 May 2025)
- Fractl: Search isn’t dying, it’s redistributing
- Anicca: Datos/SparkToro Q4 2025 takeaways and Tubefilter coverage