Three and Four Word Searches Now Carry 60% of AI Overview Ads. Your Dealership's Best Queries Were Already That Length.

On September 11, Jason Tabeling published a set of Google Ads numbers at Search Engine Land that describe a shift in how long people's searches are, comparing January 2025 with August 2026.

The impression side moved first. One and two word queries fell from 42% of impression share to 24%. Three and four word queries climbed from 33% to 48%.

The conversion side moved further. One and two word queries fell from 62% of conversion share to 52%. Three and four word queries went from 20% to 46%. Five and six word queries tripled from 3% to 9%. Seven word and longer queries went from 1% to 4%.

Those are large numbers and they are being passed around this week with a short list of instructions attached: double down on the long tail, master hyper-specific ad scent, aggressively realign bidding and budgets.

Two of those three are fine. The first one points at the wrong end of the distribution, and a dealership that follows it will spend money in the emptiest part of the auction.

Read the sourcing before you read the chart

The article never says whose accounts these numbers come from. There is no sample size, no vertical, no geography, no account count. It is presented as aggregate trend data with no methodology attached.

That does not make it wrong. Tabeling is Head of Solutions at Further and knows the data he is looking at. It does make it a direction rather than a benchmark. Nobody should be walking into a Monday meeting saying "industry conversion share for short queries is 52%," because the article does not support that sentence and neither does anything else published this week.

What makes the direction credible is that two independent datasets point the same way.

Google published its own figure on May 19 in a post about a year of AI Mode in the US, written by Shivani Mohan, its VP of Data Science and UXR. The line is plain: "the average AI Mode search is triple the length of a traditional Search query." The same post says AI Mode passed a billion monthly active users globally and that AI Mode queries have more than doubled every quarter since launch. Google has an obvious interest in AI Mode sounding transformative, so weigh it accordingly, but a stated multiple from the company that owns the log files is worth more than an inference from the outside.

Then there is Adthena, which tracks paid search across 29.1 million queries covering 10 or more industries and three markets, US, EMEA and APAC, between November 2025 and March 2026. Its reporting on query length carries the line that should change what a dealer does on Monday: "Three to four word searches account for 60%+ of all AI Overview ad appearances." And then the sentence nobody is quoting: "Long-tail queries, where AIOs are most naturally triggered, are largely uncontested."

So the ad inventory is concentrated in the mid-tail. The genuinely long conversational strings trigger AI answers constantly and almost nobody is bidding against you there.

Why the long tail advice is backwards

There is an older version of this mistake in auto retail and most of us lived through it.

When "near me" searches took off, a generation of dealer accounts got rebuilt around the phrase itself. Stores bought "car dealer near me" and "used cars near me" and put "near me" in the headline of every ad. The behavior underneath the phrase was real. The instruction drawn from it was too literal. The winning move was structural, which was getting the location, hours, inventory and proximity signals right, so that you qualified for local intent no matter how the shopper phrased it.

This is the same shape. A distribution is shifting toward longer strings. The literal instruction is to go buy longer strings. The useful instruction is to look at where the money and the competition actually sit in the new distribution, which is three and four words, and then decide whether your account can even reach the rest of it.

There is also a second reading of the query length shift that nobody is discussing, and a dealer should hold it in mind. Query length distribution in a Google Ads account describes the queries that reached the ads auction. Google spent the last two years changing which searches produce a traditional results page at all. Some of this chart could be shoppers typing differently. Some of it could be Google routing the short stuff elsewhere and handing the auction a different mix. Both produce the same chart, and they call for different responses. Nobody outside Google can currently tell them apart.

What this means for your dealership

You already live in the mid-tail. Go read your own search terms report. "toyota dealer near me" is four words. "used silverado for sale" is four words. "honda crv lease deals" is four words. Franchised dealers have been winning three and four word queries since before AI Mode existed. A chart showing conversion share concentrating there is describing the queries you already convert.

The bucket that lost share is where your brand and model terms live. One and two word searches are "toyota," "silverado," "ford dealership." That is the part of a dealer budget most likely to be defended by tight exact match and years of accumulated settings. If that bucket really is shrinking as a share of conversions, the account structure protecting it is protecting a smaller and smaller pot.

Your negative keyword list was built for a one and two word world. This is the expensive one. Most franchised dealer accounts carry negative lists assembled over five or more years, full of blocks on words like how, best, worth, vs, compare, review, problems and reliability. Those words are exactly what makes a query four or six words long. A negative list built to keep the budget tight in 2019 is now a filter that excludes the growing part of the distribution, and it does it silently, because blocked queries do not appear in your reports as a loss.

The uncontested long tail is real inventory you probably cannot reach. If long conversational queries are largely uncontested, that is cheap traffic. Reaching it means broad match, AI Max or Performance Max, which is the same fence we wrote about when Google confirmed exact and phrase match can now serve inside AI Mode. Getting to the uncontested part of the auction means loosening the controls that most dealer accounts were deliberately tightened to keep.

Landing pages have to match a longer string. A four word query carries a filter inside it. "used tacoma under 30000" is a price cap. "certified crv all wheel drive" is a trim. Sending that shopper to a homepage or an unfiltered inventory page throws away the specificity that earned the click. This is the "ad scent" item on Tabeling's list and it is the one worth taking literally.

Do not read a paid query shift as your organic AI problem. These are separate surfaces with separate symptoms. We covered why rising click rate on shrinking impressions misleads dealers and why the AI report in your Search Console counts impressions and nothing else. Diagnose them separately or you will spend money fixing the wrong one.

What to do about it this week

1. Run the same cut in your own account. Pull 24 months of search terms, bucket every query by word count, and chart impression share and conversion share by bucket. This is the only version of Tabeling's chart that is actually about your store. It takes an afternoon and it settles the question.
2. Audit your negative keyword list against the mid-tail. Sort by date added. Anything older than three years, check what it blocks today. Question words and comparison words are the ones to look at first.
3. Check what share of your conversions already come from three and four word queries. If it is high, the headline shift is not a threat to you and you can stop reading the panic version of this story.
4. Pick one campaign to test looser matching on. Not the brand campaign. Use a used inventory or a model line campaign where a bad match costs the least, and see whether the uncontested long tail shows up as cheap conversions or as waste.
5. Fix the landing page mapping for your top 20 mid-tail queries. Each one should land on a filtered SRP that already reflects the filter inside the query.
6. Ask your agency for the word count breakdown by name. If they cannot produce it from your search terms report, they are not looking at the thing the entire industry is currently arguing about.

The part worth watching

The number to track is whether the long tail stays uncontested. Adthena's window closed in March. If dealers and their competitors move into those queries over the next two quarters, the cheap inventory stops being cheap and the correct answer flips back toward defending the mid-tail you already own.

Until someone publishes an auto-specific cut of this with a stated methodology, every number in this story is cross-industry. Your own search terms report is the only auto retail dataset any of us can actually read, and it is sitting in your account right now.

If you want a second set of eyes on what your query mix has been doing for the last two years, and on what your negative keyword list is quietly costing you, book a time with us and we will go through the account together.