Long-tail keywords: what they are and the clicks they get

If you've ever put together a list of long-tail keywords, I bet it looked something like this: forty long phrases, several of them phrased as questions, each one with "low competition" noted next to it. All very reasonable. What that list doesn't tell you is how many clicks they bring in today, now that Google answers a lot of questions above the results.

So we went and looked at Search Console data from 199 company websites in the US, Latin America and other markets. In September 2026, long or question-style searches made up 9% of non-branded impressions (median), but only 3.5% of clicks. And that's even though they ranked higher than everything else.

My read is that long-tail keywords are still worth it, as long as you choose them and measure them by the clicks and conversions they bring. Looking at position alone will give you an overly optimistic picture.

Let's start with the basics and then get to the numbers.

What are long-tail keywords?

Long-tail keywords are search phrases that are longer and more specific than a generic term. Most guides draw the line at three or more words (some say two), although what really defines them is specificity: whoever types them already knows pretty well what they need.

The name comes from the shape of the demand curve. A handful of short terms, called short-tail or head terms, each concentrate a lot of volume. Behind them comes a very long tail of queries that are rarely repeated, but add up to a huge number. In Spanish they're also called "palabras clave de cola larga", which sounds odd, so in practice everyone just says long tail.

One example from each side, so you can see the difference:

The last example is real. On our own blog, the Spanish query "que es una matriz de análisis en investigación" brought 68 impressions and 5 clicks in three months to the analysis matrix guide, with an average position of 4.4. Low volume, but a person who arrives knowing exactly what they want.

One detail to watch: long and specific don't always go together. "What is insurance and what is it for" has seven words and is a pretty generic question, the kind Google answers on its own. That difference will matter further down. If you want a refresher on keyword types from scratch, there's the guide on what keywords are.

Why do people say long-tail keywords convert better?

The guides that rank on Google for this topic today repeat the same advantages, and they're mostly right.

• Less competition. Few pages target that exact combination of words, so it's easier to show up in the top results on Google. For a new site with little authority, it's usually the most realistic way to start.

• Clearer intent. Someone searching for "outdoor plastic roller blinds" has probably moved past the browsing stage. The user's intent is written right into what they typed.

• Better conversion. The traffic you get is smaller but more qualified, and a bigger share of it ends up as a lead or a sale.

• Lower cost in Google Ads. In paid search, long-tail keywords usually have a lower cost per click because fewer advertisers are bidding on them.

All of that still holds. What those guides don't offer is any data on how many clicks actually arrive. The figure "70% of searches are long tail" has been going around for years, usually without a source. And for a marketing team there's a more practical question: how much do they weigh in your Search Console, and how many clicks do they leave you?

So we measured it.

How much do long searches weigh on Google?

First, the definition we used, because it's narrower than "long tail". We counted as conversational any query with seven or more words, any that starts with a question word (what, how, which, how much, where, why, when and their Spanish equivalents) or any that includes a question mark. It's the slice of the long tail that looks most like the way we talk to ChatGPT.

With that definition, in September 2026 conversational queries made up 9.0% of non-branded impressions and 3.5% of clicks (medians across 199 sites). There's a lot of variation: on a quarter of the sites they account for less than 3.8% of impressions, and on another quarter, more than 19.8%.

Two things caught my attention.

The first is that their share barely moved in three months. Counting all visible queries, it went from 8.3% in June to 8.7% in September (median across 199 sites), and it only rose on 45% of the sites. At least on Google, the idea that "everyone searches like they're talking to ChatGPT now" doesn't show up in this period (there may be seasonality between June and September).

The second is the difference between regions. In the US, questions and long phrases weigh almost twice as much as in Latin America (13.8% versus 7.0% of impressions). How long people write depends on the language, so take this as a rough guide. What does repeat in both regions is the gap between impressions and clicks.

Why do long searches get fewer clicks if they rank higher?

This is the most striking part of the study. Conversational queries show up higher than the rest (average position 8.5 versus 10.8, better on 66% of the sites) and yet their CTR is less than half: 0.26% versus 0.67%. On the typical site, the rest of the searches get 1.98 times more clicks per impression (median across 166 sites), and on 73% of the 176 sites with at least 500 conversational impressions, their CTR is worse.

If it were only about position, it should be the other way around. So the click gap isn't explained by ranking lower in Google's results. The gap shows up in the US (0.24% versus 0.70%) and in Latin America (0.58% versus 1.10%).

One possible explanation is that for questions, Google answers on the results page itself, with featured snippets, AI Overviews or AI Mode, and the person takes the answer without clicking. Since BERT, in 2019, Google understands long phrases much better, and with AI Overviews it also answers them. But careful: this data can't prove that explanation. Search Console mixes AI Overviews and AI Mode traffic in with the rest of web search, so all we see is the correlation.

The practical consequence for marketing is clear, though. An informational long-tail keyword on page one can look great in the rankings report and bring very little to the site. If your content strategy is measured in positions, these queries will inflate your report. I wrote more about how to read this in what CTR is and how to improve it, and about how to show up in those answers in the guide to Google's AI Mode.

How do you find long-tail keywords?

The best source is your own Search Console, because it shows the real searches people use to reach your site. This is the method we used in the study, adapted for a single site:

1. In Search Console, go to Performance and open the queries tab. Filter to the last three months.

2. Add a query filter with "Custom (regex)" and paste ^(what|how|which|why|when|where|who|can|does|do|is|are|should) to see the questions (for Spanish queries, use the question words in Spanish). Search Console uses RE2 syntax.

3. Repeat with ^([^ ]+ ){6,}[^ ]+$ to see queries with seven or more words, and with \? for those that include a question mark.

4. Remove branded queries from the list and compare the CTR of that group with the rest of the site.

That's where you'll find the long-tail terms already bringing you impressions. For the ones you don't have yet, there are four more sources:

• Google's autocomplete, which suggests real searches as you type, and the "People also ask" box on the results page.

• SEO tools like Google Ads' Keyword Planner or Ahrefs, filtering by length or by question format. They give you estimated volume, which in the long tail is usually small. (This very search, "palabras clave long tail" in Spanish, gets about 150 searches a month in Mexico according to Ahrefs and almost none in Chile.)

• The competition: which long phrases bring visits to the sites that beat you on Google.

• Sales and customer service: the questions that come in through chat or email are long tail in its purest form, and you already know they come from people who buy.

How do you use long-tail keywords in your content?

With the data above, I'd change the selection criteria. The long-tail terms Google can answer in two lines ("what is", "how tall is", "when is") will give you impressions and few clicks. The ones that need your site to be resolved have a better chance of bringing visits: comparisons, prices, conditions, quotes or cases from your industry. That's my hypothesis and not a result of the study, but it's the cheapest way to organize your content strategy.

Then, when writing:

1. One or two long-tail keywords per page, covered in depth, with their close variants on the same page. One URL per variant ends up with pages competing against each other.

2. Check the intent before writing. If the search is informational, the page has to teach; if it's commercial, it has to help people decide. Mixing them is a common mistake.

3. Put it in the title, a subheading and the first paragraph, naturally. Forcing it into every paragraph reads badly and doesn't help.

4. Answer the question in the first two sentences of the section and then expand. If Google or ChatGPT are going to cite someone, make it a passage of yours that answers completely. I covered this in detail in how ChatGPT rewrites your questions and in the SEO writing guide.

5. Skip what nobody searches for. A long tail invented by the team, with no impressions or volume, ends up being just a long title.

And measure differently. For this kind of keyword, instead of position, look at clicks, leads and whether you're cited in AI answers. The AI report in Search Console already includes impressions from AI Overviews and AI Mode.

Methodology and limitations of the study

The data comes from Google Search Console (web search) for 222 company domains in the US, Latin America and other markets that we have access to. The criteria were set before looking at the results: one property per domain, excluding unverified properties and our own sites. Sites with 1,000 or more impressions in each period were included: 199 sites (117 in the US, 39 in Latin America and 43 in other markets, based on the country with the most impressions).

The main period runs from September 6 to October 3, 2026, and the comparison period from June 7 to July 4, 2026 (28 days each). A search counts as conversational if it has seven or more words, starts with a question word in Spanish or English, or contains a question mark. We excluded queries containing the domain name to approximate non-branded searches. For each site we calculated the share of impressions and clicks, the CTR and the average position of each group, and then medians and 25th and 75th percentiles across sites. CTR was compared on sites with 500 or more non-branded conversational impressions.

The limitations, no fine print:

• Search Console hides infrequent queries for privacy reasons, and long ones are the most affected. Their real share is higher than measured.

• The brand filter is an approximation based on the domain name and doesn't catch variants.

• The definition of a conversational search is operational, and English and Spanish aren't written at the same length.

• AI Overviews and AI Mode traffic is included in web search and can't be separated.

• This is correlation, not causation. It covers two 28-day windows with possible seasonality, and it's a convenience sample: it doesn't represent the market.

Frequently asked questions about long-tail keywords

How many words does a long-tail keyword have?

Guides usually say three or more, but the number is a convention. What defines it is being specific and having low individual volume. In the study we used seven or more words, or question format, to isolate the most conversational ones.

Do long-tail keywords work in Google Ads?

Yes. They usually have a lower cost per click because fewer advertisers compete for them, and the clearer intent helps sell. Still, check that they have enough volume for the campaign to learn.

Are ChatGPT-style searches growing on Google?

In our data, between June and September 2026 their share went from 8.3% to 8.7% of impressions (median across 199 sites). It's a short period and there may be seasonality, so it's too early to call it a trend.

What I'd do this week

Open Search Console, paste the question filter from step 2 and write down the ten questions with the most impressions. Next to each one, note its CTR and whether Google already answers it above the results (search it yourself, it takes a minute per query).

For the ones with lots of impressions and almost zero clicks, what you want is to be the source cited in the answer. The ones that ask people to compare, get a quote or decide are the ones that deserve their own page. And in your next report, for your long-tail keywords, show clicks and conversions next to position. (That way nobody gets excited about a top 3 that brings in no one.)