Over the past week, Google Trends gave ChatGPT an average worldwide search interest of 69, against 41 for Gemini, 20 for Claude, 2 for Grok and about 0 for Cursor AI, on a scale where 100 is the highest point in the period. By that measure, people search for ChatGPT roughly three and a half times as often as they search for Claude. If you want your site to appear in these AI answers, the main work is ordinary search work: make your pages crawlable, indexable and easy to quote. Google says there is no special markup for its AI features, and a 2024 research paper found that adding citations, quotations and statistics helped far more than keyword stuffing.

These numbers come from Google Trends, checked on November 9, 2026, using the term comparison ChatGPT, Claude, Gemini, Grok and Cursor AI, worldwide, past seven days, web search. Below are the numbers, what they can and can't tell you, and a practical checklist for showing up in AI answers.

Which AI do people search for most?

Search termAverage interest, past 7 days
ChatGPT69
Gemini41
Claude20
Grok2
Cursor AI0

The numbers are relative. Google Trends scales each comparison so the highest point in the chart is 100, so 69 does not mean 69 million searches or 69 percent of anything. It is useful for comparing terms within one chart, and it can't tell you absolute volume. Cursor AI's 0 means it was too small to register against ChatGPT in the same chart, and doesn't mean no one searched for it.

Two more cautions apply to this particular comparison. Gemini is also a zodiac sign and Claude is a common first name, and this comparison used search terms instead of Google's disambiguated topics, so those two may be slightly inflated by unrelated searches. People who use an AI tool every day also tend to open the app directly, so search interest is closer to public curiosity than to actual usage. A tool can gain heavy users while its Trends line stays flat.

The regional view for ChatGPT is also interesting. Its top five regions by search interest for the week were Rwanda, Iran, Tanzania, South Africa and Qatar. Trends measures share of searches within each region, so small countries with concentrated interest can outrank much larger markets.

What do SEO, AEO and GEO mean?

These three terms overlap, and people use them loosely. A working set of definitions:

  • SEO, or search engine optimization, helps your pages rank and earn clicks in search results.
  • AEO, or answer engine optimization, shapes content so a system can lift a direct answer from it, such as a featured snippet or a voice result.
  • GEO, or generative engine optimization, aims to get your content used and cited inside AI-written answers such as AI Overviews, ChatGPT search or Perplexity. The term was popularized by the research paper discussed below.

In practice these aren't separate jobs. Google's documentation on AI features describes eligibility in plain SEO terms. A page has to be indexed and able to appear in Search with a snippet, and even then inclusion isn't guaranteed.

Do you need special markup or llms.txt for AI search?

For Google, no. Its guidance says, "There's also no special schema.org structured data that you need to add." It also says you don't need new machine-readable files or AI text files to appear in AI Overviews or AI Mode. Structured data is still useful for other reasons, and Google asks that it match the visible text on the page.

Other AI products work differently, so check each one that matters to you. OpenAI, for example, documents separate crawlers. OAI-SearchBot is used to surface sites in ChatGPT's search features, GPTBot crawls content that may be used for training, and ChatGPT-User visits a page when a person's request triggers it. You can allow the first while blocking the second, since the two settings are independent. OpenAI says that sites opted out of OAI-SearchBot won't be shown in ChatGPT search answers, though they can still appear as navigational links, and that robots.txt changes can take about 24 hours to affect search results.

What actually improves your chances of being cited?

The clearest public evidence on this is the paper "GEO: Generative Engine Optimization" by Aggarwal and colleagues, presented at KDD 2024. The researchers tested content changes against a benchmark of queries and measured how visible a source became in generated answers. Their results:

MethodChange in position-adjusted word count
Adding quotations+41%
Adding statistics+30.6%
Citing sources+27.5%
Keyword stuffingAbout 8.3% worse than baseline

That result needs careful reading. It's a 2024 study of test engines, and real products change their systems often. The authors also found that the best method depends on the subject: an authoritative tone worked best for debate and history queries, statistics for law and government, and quotations for people and society. Treat the table as a direction to test and not as a formula. It does fit common sense, though. An answer engine needs something specific to quote, and a page full of vague claims gives it nothing.

A checklist for getting found in AI answers

Work through these in order, because each step depends on the one before it.

  1. Make sure search engines can index the page. Look it up in Google Search Console's URL Inspection tool and fix any blocked or noindex status. Google's first requirement for AI features is the same as for normal search.
  2. Check crawler access. Confirm that robots.txt and any CDN or firewall rules allow the bots you care about. A bot-protection setting on a CDN can block crawlers without your noticing, and Google specifically lists CDN and hosting rules in its advice.
  3. Put the answer first. Start the page with two or three sentences that answer the main question fully, including the key number or name. A reader and a retrieval system both benefit.
  4. Use real questions as headings. "How much does it cost?" maps to a query better than "Pricing considerations." Follow each heading with a direct answer in its first sentence.
  5. Keep key content in plain text. Google asks that important content be available in textual form. Facts that appear only in an image or a video are harder for any system to use.
  6. Add evidence. Include specific figures and dates, name who said what, and link to the original source. This is where the paper's results point, and it also makes your page more trustworthy to people.
  7. State what you don't know. A clear line such as "the report doesn't give a date" can be more useful than a confident guess, and it protects your credibility if the facts change.
  8. Show dates and keep pages current. Mark when the information was checked, and update important pages when the facts move.
  9. Write for one reader and one problem. A focused page is easier to match to a question than a broad page that touches everything.

Check which AI crawlers can reach your page

This short script reads your site's robots.txt and reports whether common crawlers may fetch a URL. It uses only Python's standard library.

#!/usr/bin/env python3
"""Check whether common search and AI crawlers may fetch a URL under robots.txt."""
import sys
from urllib.parse import urlsplit
from urllib.robotparser import RobotFileParser

BOTS = ["Googlebot", "Bingbot", "OAI-SearchBot", "GPTBot", "ClaudeBot", "PerplexityBot"]
url = sys.argv[1]
parts = urlsplit(url)
robots = RobotFileParser(f"{parts.scheme}://{parts.netloc}/robots.txt")
robots.read()
for bot in BOTS:
    print(f"{bot:<16} {'allowed' if robots.can_fetch(bot, url) else 'blocked'}")

Run it with python3 ai_bots.py https://example.com/page. It was tested against a local server with a robots.txt that blocked GPTBot, and it reported that correctly. It only reads robots.txt, so it can't detect a block at your CDN or firewall. Check your server logs for those crawler names to see what really reaches you.

Mistakes that waste your time

The advice around AI search moves quickly, and a lot of it comes from people selling a checklist. Before you act on a tip, ask whether a primary source supports it. Adding an llms.txt file because someone told you it's required for Google is a good example of effort that Google's own guide says you can skip.

Publishing large numbers of nearly identical pages in the hope of matching every prompt is riskier. Google's spam policies describe scaled content abuse, which covers many pages made mainly to rank and offering little value to readers, and it can affect a whole site. One thorough page that answers a real question is safer and usually more useful than twenty thin variations.

Keeping your best facts inside a PDF, a screenshot or a chart with no surrounding text also makes them harder to reach. Put the key number in a sentence near the chart. The same goes for content that only appears after a script runs or a login opens, since a crawler may never see it. Stale numbers are a quieter problem. An answer engine that finds a 2023 figure on your page may quote it as current, so date your statistics and review the pages that bring you the most traffic.

How to measure whether it's working

Google says traffic from AI features is included in the overall figures for the Web search type in Search Console's Performance report. There is no separate line for it there, so watch the totals alongside your individual pages, and expect the numbers to take time.

For AI products outside Google, look at referral traffic in your analytics for domains such as chatgpt.com and perplexity.ai. Referrals only capture people who click through, and many readers get the answer and stop there. To fill that gap, keep a simple monthly record: pick ten questions your audience asks, ask each AI tool, and note whether your site is cited. It's a small sample and results can vary from one try to the next, so treat it as a trend line. Keep impressions, visits, citations and purchases as separate numbers, because each can move without the others.

What these Trends numbers mean for your content plan

ChatGPT still commands the most public curiosity, so comparison and how-to content that uses its name can reach a wide audience. Claude's and Gemini's numbers are much smaller on this measure, although they are still large in absolute terms, and developer audiences may be more concentrated around them. Grok's and Cursor's low numbers show how much narrower these tools' search audiences are, and a focused article on one of those can still do well because there's less competition.

If you plan by search volume alone, you will keep writing the same articles as everyone else. A better use of Trends is to find the specific question within a broad topic, then answer that question better than the existing pages do. The related searches attached to a trend are often more useful than the headline term.

If you want this as a repeatable system, Search Visibility Lab is a practical SEO, GEO and AEO playbook: a 92-page e-book and 22-page working kit with a 30-day plan for $19, or $49 with 150 AI prompts and an Excel planner. It says plainly that it can't promise rankings or AI citations.

Frequently asked questions

Is ChatGPT more popular than Gemini and Claude?

In this week's Google Trends comparison, ChatGPT led with 69, ahead of Gemini at 41 and Claude at 20. That measures search interest, which isn't the same as active users.

What is the difference between AEO and GEO?

AEO focuses on being the direct answer in a result such as a featured snippet. GEO focuses on being used and cited inside AI-generated answers. Most of the practical work overlaps.

Do I need llms.txt to appear in Google's AI Overviews?

No. Google says you don't need new machine-readable files or AI text files for its AI features.

Should I block GPTBot?

That is a decision about training use. OpenAI treats GPTBot and OAI-SearchBot separately, so you can block training while still allowing your site to appear in ChatGPT search.

Does keyword stuffing work for AI search?

In the 2024 GEO study it performed worse than leaving the content alone.

Sources