Long-tail queries are search queries that are individually rare - typed a handful of times a month, or never before - but that collectively account for a large share of everything people ask a search engine. "Shoes" is a head query. "Waterproof trail running shoes for wide feet under 150" is a long-tail one. The idea exists because search demand is not evenly spread: a small number of phrases attract enormous volume, while millions of others form a thin, extended "tail" on the chart. For advertisers and publishers, that tail is where competition is lighter and intent tends to be more precise, and also where data is scarcest.

The term describes a position on a distribution, not a length. Longer phrases usually sit further down the curve, which is why the two ideas are often conflated, but a one-word query can be long-tail and a five-word query can draw hundreds of thousands of searches a month.

Where the term comes from

The statistics are old: long-tailed frequency distributions were studied from at least 1946. In February 2003, Clay Shirky's essay "Power Laws, Weblogs and Inequality" described how a few blogs collect most links while millions share very few.

The business usage came from Chris Anderson, then editor of Wired, whose article "The Long Tail" appeared in the magazine's October 2004 issue. Anderson argued that online retailers such as Amazon and Netflix could earn more in aggregate from niches than from hits, because digital shelves cost almost nothing. He expanded the argument into a book, published by Hyperion on July 11, 2006.

Search marketers adopted the vocabulary quickly. A December 15, 2005 piece in Chief Marketer applied it to keywords, citing Excite co-founder Joe Kraus's observation that about 97% of queries the engine saw in 2000 were unique, while advertisers concentrated on the top 3%. Google supplied its own numbers. In 2007, Udi Manber, then Google's vice-president of engineering, said 20 to 25% of the queries Google saw each day had never been seen before, according to Search Engine Land, which later described the figure as a "ballpark estimate". By 2013 Google had revised the share to 15%, a number it repeated in April 2017 and again when announcing BERT in 2019. In March 2025, Google's John Mueller told Search Central Live in New York that the figure is "always around 15%" when recalculated, adding that he had expected large language models to push it higher, according to Search Engine Journal.

How the long tail is measured

There is no official threshold. Ahrefs, whose long-tail guide was last updated on May 27, 2026, defines long-tail keywords purely by search volume and says there is "no specific search volume threshold". Its own US database illustrates the shape of the curve: keywords with fewer than 10 searches a month account for almost 93% of the database, roughly 2.3 billion phrases, while just under 18,000 keywords exceed 100,000 monthly searches.

The company splits the tail into two kinds. A supporting long-tail query is a less popular variant of a broader topic - a page ranking for "healthiest dog treats" will usually also rank for the variants. A topical long-tail query is itself the most common way of asking for something narrow, so it needs its own page. The distinction matters because the first is captured almost automatically, while the second requires dedicated content or a dedicated keyword.

The data is incomplete. Anything below a tool's reporting floor appears as zero, and queries never typed before cannot appear in any planning tool, so roughly 15% of daily demand is, by definition, unplannable.

In paid search, the advertiser bids on keywords and the user types a query, also called a search term. Match types decide how far apart the two may be. An exact match keyword reaches only the same query or close variants. Phrase match reaches queries that include the keyword's meaning. Broad match reaches anything Google considers related. The long tail is mostly reached through looser matching, because no advertiser can list every rare variant.

Google's systems treat rare keywords specially. A keyword with "very little to no search history" receives a low search volume status in Google Ads and stops triggering ads, according to Google's help centre. The system evaluates worldwide searches over the previous 12 months and reactivates the keyword if volume rises "even a small amount". Google publishes no numeric cut-off.

Visibility after the click is also limited. Since September 2020, the search terms report has only shown terms that "a significant number of users searched for", according to Google's statement at the time. Long-tail queries are the ones most likely to fall below that line, so advertisers pay for clicks on queries they cannot see. LinkedIn research cited in May 2026 put the share of search term data hidden by these thresholds at between 27% and 73%.

Why advertisers and publishers care

The traditional case for the tail rests on cost and intent. Fewer bidders compete for rare phrases, so cost per click is usually lower, and a specific query tends to come from someone further along a purchase. Mike Ryan, head of ecommerce insights at Smarter Ecommerce (smec), said in October 2026 that earlier smec work on Shopping query sculpting had found long-tail queries showed strong intent and cheaper CPCs, while noting he had not repeated the analysis "in a long time".

In organic search the logic is similar. SISTRIX's Visibility Index weights rankings by volume, so a top position for a long-tail phrase counts for less than a lower position on a head term, yet sites built on thousands of niche pages can collect substantial traffic in aggregate.

AI search is changing the input. In April 2026, Google's head of Search, Liz Reid, said people now give the system "their actual need" rather than translating it into keywords. Google's own May 19, 2026 report on AI Mode said the average query there is triple the length of a traditional search. Longer, conversational prompts are almost all unique, which pushes more demand into the tail.

AI Max and the keywordless tail

Google's answer for advertisers is to stop asking them to enumerate the tail. AI Max for Search campaigns, announced on May 6, 2025 by Brian Burdick, senior director of product management, adds "search term matching", which expands existing keywords with broad match and keywordless technology. Google said advertisers activating it typically saw 14% more conversions or conversion value at a similar cost per acquisition or return on ad spend (ROAS), rising to 27% for campaigns still relying mainly on exact and phrase keywords. Those figures come from Google's internal 2025 data, not an independent audit.

The product has since become the default route into the tail. Google began upgrading campaigns using the broad match setting to AI Max from September 1, 2026, and Microsoft Advertising retired Predictive Matching in favour of its own AI Max. Smec data covering 383 million impressions found that by July 2026 more than 25% of exact match impressions had been broadened by AI Max, reaching about 29%.

Limitations and disputes

The most persistent dispute concerns size. A widely repeated "search demand curve" claims about 70% of searches are long-tail. The figure appears without a primary dataset in much SEO writing, including a Search Engine Land explainer by Anthony Atkins, and the share depends entirely on where the head ends. Ahrefs' 93% figure counts keywords, not searches, which is a different measure.

Whether the tail performs is also contested once automation chooses it. In November 2025, smec found AI Max conversions arrived at roughly 35% lower ROAS than other match types across more than 250 retail campaigns. Xavier Mantica's four-month test put AI Max cost per conversion at $100.37 against $43.97 for phrase match, and an August 2025 test by Ezra Sackett of Monks found 99% of AI Max impressions produced no conversion across about 30,000 search terms.

Reporting is a third problem. For searches from Lens, AI Mode, AI Overviews and autocomplete, Google says the reported term "represents the best approximation of the user's intent", so the long tail shown in the report may not be what anyone typed. On October 25, 2026, PPC Land reported that the search terms report had gained four undocumented source columns labelling traffic from AI Max expanded matches.

The original economics are disputed too. Anita Elberse argued in Harvard Business Review in 2008 that online markets magnify hits rather than niches, a challenge Anderson rebutted.

Long-tail keywords. Keywords are what an advertiser or SEO targets; queries are what users type. Long-tail keywords are an attempt to anticipate long-tail queries, and the two lists never fully overlap.

Low search volume. A Google Ads keyword status, not a category of query. It switches off rare keywords while the queries themselves can still be matched by broader ones.

Long-tail inventory. In programmatic advertising, the "long tail" refers to thousands of small publishers selling impressions through exchanges. It shares the statistical shape, not the subject.

Conversational queries. Natural-language prompts in AI Mode or chatbots are usually long-tail by volume, but long-tail queries also include short, rare searches such as part numbers.

Recent developments

Google's AI Max migration removed the campaign-level broad match setting in August 2026, and the Dynamic Search Ads automigration was pushed to February 2027. AI Overviews now appear on 88% of tracked commercial keywords, according to Semrush data reported this month, changing how many clicks any organic ranking, head or tail, can deliver. The AI Max source columns reported today suggest the question has shifted from which long-tail query triggered an ad to which system chose it.

Timeline

  • 1946 - Brown and Tukey's paper, the earliest "long tail" use recorded in MathSciNet.
  • 2000 - About 97% of queries at Excite are unique, per co-founder Joe Kraus.
  • February 2003 - Clay Shirky publishes "Power Laws, Weblogs and Inequality".
  • October 2004 - Chris Anderson's "The Long Tail" appears in Wired.
  • December 15, 2005 - Chief Marketer applies the long tail to search keywords.
  • July 11, 2006 - Anderson's book "The Long Tail" published by Hyperion.
  • 2007 - Udi Manber says 20 to 25% of Google queries are new each day.
  • 2008 - Anita Elberse challenges the theory in Harvard Business Review.
  • 2013 - Google first cites the 15% new-query figure.
  • April 25, 2017 - Google restates that 15% of daily searches are new.
  • September 2020 - Google limits the search terms report to terms searched by "a significant number of users".
  • November 25, 2021 - Ahrefs publishes its long-tail keywords guide.
  • March 2025 - John Mueller says the new-query share is "always around 15%".
  • May 6, 2025 - Google announces AI Max for Search campaigns.
  • November 6, 2025 - Smec and Xavier Mantica publish AI Max performance tests.
  • May 13, 2026 - Google's "best approximation of the user's intent" wording reported.
  • May 19, 2026 - Google reports AI Mode queries average three times traditional length.
  • July 2026 - Smec data shows about 29% of exact match impressions broadened by AI Max.
  • August 19, 2026 - Microsoft rolls out AI Max to all accounts globally.
  • September 1, 2026 - Google begins upgrading broad match setting campaigns to AI Max.
  • October 25, 2026 - AI Max source columns reported in Google's search terms report.
  • February 2027 - Scheduled Dynamic Search Ads automigration to AI Max.

Summary

Who: Searchers generate long-tail queries; search engines such as Google and Microsoft Bing match them to ads; advertisers, SEO teams and publishers target them; vendors such as Ahrefs and smec measure them. Chris Anderson popularised the business term.

What: Individually rare search queries, often specific and frequently never seen before, which together make up a large share of total search demand. They are defined by low volume rather than word count.

When: The statistical idea dates to the 1940s; Anderson's Wired article appeared in October 2004; search marketers applied it to keywords by 2005; AI Max, from May 2025, made automated matching the main route into the tail.

Where: In organic results and in paid search auctions on Google, Microsoft Advertising and AI search surfaces such as AI Mode, visible to advertisers through search terms reports that hide many of the rarest queries.

Why: Long-tail queries carry precise intent and face less competition, but they are too numerous and too new to list one by one. That gap is the reason platforms now sell automated matching to reach them, and the reason advertisers dispute how well it works.