53 Developer Tool AI Search Statistics
Which pages AI engines cite for developer tools, what vendor docs return without JavaScript, and how to pick the next page to fix.
On this page
- Highlights
- Which Page Types AI Engines Cite for Developer Tools
- How Developers Check an AI Answer Before Using It
- How Cited Page Types Differ by AI Engine
- Which Page on a Vendor’s Own Site Gets Cited Most
- What the Articles AI Engines Cite Have in Common
- How Often AI Answers Cite Roundups and Comparison Pages
- Where Developers’ Discussions Get Cited
- How to Choose the Next Page to Fix
- Frequently Asked Questions
Empact Newsletter
One email a month on what is working in AI search and SaaS go-to-market.
Landing pages were the source of 2.6% of the citation records AI engines produced for developer-tool questions in DevTune’s April to June 2026 count.1 A record is one source link in one answer. Blog posts were 34.3% of the records and documentation 17.3%.1 Only video, at 1%, was cited less often than the landing page, so a launch plan that funds the landing page first starts with one of the page types engines drew on least.
Citations are not sales, and a landing page does work an AI answer never measures. The share shows which kinds of page the engines drew on overall, and it cannot say how often any one landing page gets cited.
Empact Partners reads a developer tool’s docs and blog before its landing page when working on AI search visibility, so that ChatGPT, Google AI Mode and Claude cite and recommend the product. Shares like these tell us which kinds of page to read first.
Highlights
- Blog posts were 34.3% of DevTune’s developer-tool citation records, documentation 17.3% and landing pages 2.6%, which means the pages explaining how a tool works supplied about half the sources.1
- 75.98% of citation records went to domains outside the 478 brands DevTune tracks. Three in four pointed away from a tracked vendor’s own site.1
- Documentation was 21.9% of ChatGPT Search’s citations and 11% of Google AI Mode’s, so your docs weigh more on some engines than on others.1
- 74.4% of pairings of an answer with a vendor’s comparison or best-tools page named the publisher, and 17.1% named a rival without the publisher. A ranked page you publish usually brings your own name into the answer.1
- 38.0% of 13,162 developers read the official docs before using an AI answer, and 9.6% use the answer as it comes, which makes the docs the place many answers get checked.2
- 61 of 380 docs front pages returned under 150 words to our request without JavaScript, leaving a reader that skips scripts little to quote.3
Read together, they say where to look before the landing page gets more money: your docs and blog, and the outside pages your category’s answers already cite.
Which Page Types AI Engines Cite for Developer Tools
DevTune, which sells AI search tracking for developer tools, asked five AI engines 1,075 test questions across 43 developer-tool categories every week from 2 April to 30 June 2026. It sorted all 126,814 citation records by the kind of page each link pointed to.1 Blog posts took 34.3%, documentation 17.3%, discussions 8.8% and landing pages 2.6%.1
Citation Records by Page Type, Developer-Tool Questions, April to June 2026
Share of 126,814 citation records by page type, five AI engines, DevTune
View the data
| Label | Value |
|---|---|
| Blog post | 34.3% |
| Documentation | 17.3% |
| Discussion | 8.8% |
| Product page | 7.5% |
| Article | 7.4% |
| Comparison | 7.3% |
| Listicle | 7.2% |
| Other | 6.7% |
| Landing page | 2.6% |
| Video | 1% |
Together, blog posts and documentation made up 51.6% of the records.1 For a team dividing a content budget, far more of the cited sources were pages that explain how a tool works than pages built to sell it. Your own site may not follow that pooled pattern, and only a count of your own cited pages will tell you.
Before any page gets rebuilt, count which of your own pages the engines already cite, and which they skip.
Here is that count for one SaaS company we work with, from the Empact Panel, which we made to ask Google AI Mode, Claude and ChatGPT a company’s buyer questions each week and log which of its pages they cite.
The card below sorts that company’s own cited pages by site section, its name replaced with “Partner”. In the 90 days to mid-November 2026 the engines cited 68 of its own pages, 47 of them blog posts and 6 landing pages, and Claude cited no landing page at all.

Those 68 pages drew 0.1% of all the citations in the answers to that company’s buyer questions, the figure the card calls citation ownership. Almost every source those answers used sat on someone else’s site. It is one company’s card, so read it as an example of the count. Wihan van Zyl, who wrote this post and is a GTM consultant at Empact Partners, would split the work between your own pages like this:
I would treat the landing page as where a convinced developer arrives, and give the explaining work to the docs and the blog. Which of those comes first depends on what your own category’s answers cite and which of your pages they skip.
Wihan van Zyl on LinkedInOpens in a new tabDevTune also split the records by domain. The 781 domains of the 478 brands it tracks took 24.02%, and the other 75.98% went to 11,754 domains outside them.1 That outside group mixes publications, communities and reference sites with vendors DevTune does not track.
Even so, three records in four pointed at a domain none of the tracked brands owns. To know which of those pages are communities, reviews or listicles your product could appear on, you have to read the pages cited for your own category.
The free agent below does that reading for you. Enter your website and it finds the pages cited in your category’s answers where your competitors appear and you do not, sorted by kind of page, with a route onto each.
How Developers Check an AI Answer Before Using It
Whether those cited pages matter depends on what a developer does with an answer. Stack Overflow’s 2026 Developer Survey collected 30,903 responses from 23 June to 5 August 2026, each question with its own base.2
Of 16,664 who said where they turn first for an answer, 69.9% put asking an AI agent among their first three moves and 8.0% posting on a public forum.2 So for most of these developers, an AI assistant is one of the first places they look when they get stuck.
Of 13,162 who described what they do with an AI answer before using it, 9.6% use it as it comes.2 Some 38.0% read the official docs first, so more than a third check the answer against documentation. Most run the code or compare it with their own codebase:
What Developers Do Before Using an AI Answer
What respondents usually do first with an AI-generated answer, n = 13,162, 2026
View the data
| Label | Value |
|---|---|
| Run it locally | 76.5% |
| Compare with the codebase | 63.9% |
| Inspect tests or security | 53% |
| Read the official docs | 38% |
| Search Stack Overflow | 12.2% |
| Use it as-is | 9.6% |
Asked about trust, 78.6% of 13,945 rated source attribution important or very important for an AI answer to a technical question, and 48.0% of 14,304 trust AI output they can verify easily.2 The survey asked how developers check technical answers. It never asked how they choose a tool, so the next step is our reading of it: a developer who checks an answer is likely to open the pages it cites, and those pages have to hold up.
How Cited Page Types Differ by AI Engine
Which pages a developer ends up checking depends on the engine they asked, because each one leans on different sources. DevTune kept its five engines apart.1
Documentation was 21.9% of ChatGPT Search’s citation records and 11% of Google AI Mode’s.1 Discussions were 19.3% on ChatGPT Search and 1.5% on Gemini Search. Best-tools listicles were 13.4% on Perplexity and 3.7% on ChatGPT Search, and blog posts ran from 18.5% on ChatGPT Search to 46.5% on Gemini Search.1
Share of Citations by Page Type, Lowest and Highest Engine
Lowest and highest share of citations across five AI engines, by page type
View the data
| Label | Range |
|---|---|
| Blog posts | 18.5% to 46.5% |
| Documentation | 11% to 21.9% |
| Discussions | 1.5% to 19.3% |
| Listicles | 3.7% to 13.4% |
| Comparison pages | 4.5% to 9.1% |
| Product pages | 5.3% to 9.6% |
ChatGPT Search spread its citations across documentation, discussions and product pages, while Perplexity and Gemini Search leaned on blog posts and listicles. So read the figures for the engines your buyers use, one engine at a time.
On Google AI Mode, 64.4% of citations sat among an answer’s first three sources, against 38% on ChatGPT Search.1 A page AI Mode cites is usually one of the first sources a reader sees, while most pages ChatGPT Search cites sit further down.
Which Page on a Vendor’s Own Site Gets Cited Most
The engine figures pool every vendor’s pages. For your own site, the question is which section the engines cite most, and DevTune publishes that page type for each brand it tracks. By our count, documentation led on 155 of 421 vendor domains (36.8%) and a blog post on 122 (29.0%). A landing page led on 19, and 29 domains (6.9%) had no own page cited.1
Most-Cited Own Page Type on Each Vendor’s Domain
421 developer-tool vendor domains, DevTune brand table, April to June 2026
Vendor domains
View the data
| Label | Value |
|---|---|
| Documentation | 155 |
| Blog post | 122 |
| Another page type | 115 |
| No own page cited | 29 |
On that evidence, a vendor auditing its own site should check the docs first, since they led on more domains than any other page type. In API development and management, tools for the interfaces software uses to talk to other software, documentation took 39.6% of citations, the highest of DevTune’s 43 categories.1
Docs matter beyond developer tools too. Ahrefs found a recommended brand’s own documentation cited in 7.6% of ChatGPT software prompts, against under 1% for services and products.4
Where engineering owns the docs site, a docs fix needs an engineer for the rendering and a writer for what each page says. For a developer audience the docs do much of the explaining, so each front page should answer a buying question on its own. The next question is what those front pages send to a reader today.
What Vendor Documentation Returns to a Request Without JavaScript
On 13 October 2026 we requested every vendor site in DevTune’s benchmark, 421 domains, once each and without running JavaScript. Each request received only the HTML, the page markup the server sends before any script runs.3 The count records what each site returned that day. It cannot show what an engine read, and DevTune’s citations came months earlier. Of the 421 sites, 411 answered, and this is what they returned.
Only 5 of 411 sites (1.2%) shut at least one of seven named AI crawlers out of the whole site in robots.txt, the file of crawler instructions.3 So blocking a whole site is rare. Whether a docs path is closed to a crawler is a separate check, and this count does not cover it.
We opened 20 of the 61 near-empty pages in a browser that runs scripts on 13 November 2026, and requested them again without JavaScript the same day. Of the 18 still under 150 words to the plain request, 11 passed 150 words once their scripts ran, and the other 7 were short even in a browser.3 Of the 61, 16 also returned a Markdown version at the address plus “.md”, a readable copy one address away.3
So for most of the pages we rendered, the text is there, but the HTML a plain request receives leaves it out, as in the drawing below.

Where the text exists and the HTML leaves it out, the fix is an engineering one. Sending the text in the server’s HTML gives a request that skips scripts the same steps a browser shows, and a Markdown version gives it a second copy if it asks for that address.
We also compared the quarter of vendors DevTune saw cited most with the quarter it saw cited least, 102 sites each. The most-cited quarter published an llms.txt, a file that points AI systems to a site’s content, more often, 88.2% against 65.7%. It also returned docs as Markdown more often, 57.6% against 33.7%, 57 of 99 against 30 of 89 sites with docs.3
Docs Readability in the Most-Cited and Least-Cited Vendor Quarters
Share of vendor sites, least-cited and most-cited quarters, fetched 13 October 2026
llms.txt65.7%88.2%
as Markdown33.7%57.6%
found87.3%97.1%
without JavaScript15.7%12.1%
View the data
| Label | Least-cited quarter | Most-cited quarter |
|---|---|---|
| Publishes llms.txt | 65.7% | 88.2% |
| Docs served as Markdown | 33.7% | 57.6% |
| Documentation found | 87.3% | 97.1% |
| Docs under 150 words without JavaScript | 15.7% | 12.1% |
The vendors cited most may simply keep their sites in better shape, so the gap does not show that the file earns citations. Ahrefs found up to 28% of domains using its analytics publish an llms.txt, an upper bound by its own word, and 97% of those files got no requests in May 2026.5
SE Ranking, a maker of search optimization software, found 10.13% adoption and no relation to citations.6 On this evidence the file is cheap to keep, and no study here shows it bringing citations. Wihan would begin with the near-empty pages instead:
The number I would act on first is the 61 docs front pages under 150 words. A reader that skips JavaScript gets almost nothing to quote from them, and where the text only appears after scripts run, the fix is an engineering ticket on your own site.
Wihan van Zyl on LinkedInOpens in a new tabFor the rest of your site, and for every release after this one, you can use the four-page playbook below to make sure Google AI Mode and ChatGPT can reach your buyers’ pages and read them.
What the Articles AI Engines Cite Have in Common
Blog posts were the largest page type in DevTune’s count, but no study we found describes which developer-tool articles get cited. The closest evidence comes from ChatGPT and the general web, so treat it as a guide for a developer-tool blog and check it against your own category. SE Ranking modeled ChatGPT citations for 216,524 pages across 20 niches and compared average citations per page with and without each trait:7
Scroll horizontally to see all columns when the table is wider than this page.
| Page trait | Average ChatGPT citations with the trait | Compared with |
|---|---|---|
| Updated in the past three months | 6.0 | 3.6 for older pages |
| Over 2,900 words | 5.1 | 3.2 for articles under 800 words |
| 19 or more data points | 5.4 | 2.8 for pages with few |
| An FAQ section | 3.8 | 4.1 for pages without one |
SE Ranking’s figures are associations in a general-web sample, so they do not make length a target. Recently updated pages drew more citations here. For a developer tool, an update that counts is one where the facts changed, such as a new release, a renamed setting, a new price or a rival’s move. The FAQ row is the one trait that went the other way, and Wihan reads it like this:
Pages with an FAQ section drew slightly fewer citations in SE Ranking’s sample, 3.8 against 4.1. I would keep an FAQ wherever buyers ask those questions, and judge it by whether a developer finds the answer there.
Wihan van Zyl on LinkedInOpens in a new tabIn an Ahrefs study of 16.975 million cited URLs, the pages AI assistants cited averaged 1,064 days since publication, against 1,432 days for pages ranking in Google.8 The study measured age from publication and did not track updates.
Developers weigh recency as well, with 72.6% of 12,996 Stack Overflow respondents counting recent or clearly up-to-date information among the signals that make work information trustworthy.2 The Ahrefs gap is about when pages were first published, so it says nothing about updates. The developers’ answer is the better reason to keep a cited page current.
The right first page also depends on the category. In DevTune’s testing and quality assurance (QA) category, blog posts took 46.1% of citations, the highest blog share of its 43 categories.1 So a testing tool may get more from its blog and an API tool more from its docs, and your own category’s answers show which one you resemble.
How Often AI Answers Cite Roundups and Comparison Pages
Comparison pages took 7.3% of DevTune’s citation records and best-tools listicles 7.2%, 14.5% together, about one record in seven.1 DevTune then took 668 ranked pages vendors had published and checked who was named in the answers citing them, across 2,295 page-and-answer pairings. The publisher was named in 74.4% of pairings. In 17.1%, a competitor was named and the publisher was not.1
When the cited page was a comparison page, the publisher was named 78.7% of the time, against 63.2% for best-tools listicles.1 DevTune notes that being named is not being recommended. Still, across all those ranked pages the publisher’s own name usually came into the answer, and in about one pairing in six a rival took the mention instead. A developer who checks that answer may open the page, so how it treats your rivals matters.
I would state a comparison page’s criteria at the top and give every product a real paragraph, rivals included, because a developer comparing tools checks both.
Wihan van Zyl on LinkedInOpens in a new tabIn an Ahrefs study of 750 ChatGPT prompts, best-of listicles on blogs made up 43.8% of source pages on best-of prompts.4 Of the 1,100 cited listicles carrying a date, 79.1% had been updated in 2025, the study’s year, so the listicles worth joining are mostly the current ones.4
A comparison page can go stale with any release on either side, so re-read each one against your release notes and your rivals’ public docs whenever either changes. For listicles other people publish, the way in is existing article outreach, asking the authors of the ones engines already cite to add your product.
Where Developers’ Discussions Get Cited
Discussions were 8.8% of DevTune’s records. Reddit was the most-cited independent domain in its sample, with 10,831 citations across 5,109 URLs, ahead of Medium and dev.to.1
In Stack Overflow’s survey, 69.5% of 10,741 respondents used GitHub for technology content in the past year, 68.6% Stack Overflow and 53.8% Reddit.2 And 15.9% of 16,985 visit Stack Overflow more since AI arrived, to verify answers.2 DevTune gives no citation figure for GitHub or Stack Overflow, so where developers gather and what engines cite stay separate findings.
Similarweb put GitHub at 1.62% of everything ChatGPT cited in its all-topic United States count for January and February 2026, a share across every subject.9 For developer-tool questions, Reddit is the community with a measured citation count, so look there first for threads naming your rivals. Join the threads your buyers already read, and name the product only where it answers what was asked.
How to Choose the Next Page to Fix
Each section above leaves a question to answer for your own category before you decide which page to fix next:
Four checks for your own category
- Which page types carry your category’s citations on the engines your buyers use, and whether your pages are among them
- Whether your docs answer a buying question on one page, return text without JavaScript and exist in Markdown
- When each comparison page and how-to last changed, against your latest release
- Which listicles and Reddit threads keep returning in answers, and whether they name you
On the second check, practitioners on Reddit have advice on testing a JavaScript page and on writing docs that AI readers can use:
Tips from Reddit
How practitioners check and write pages for AI readers
And the point above about diffing view-source against the rendered DOM first is the right order to do this in. Half the time you find the crawler never had the chance, because the thing you wanted read only exists after JS runs.
Every page should exist independently. No worries about repeating stuff. Term glossaries are important.
good structure, explicit examples, fewer vague links like ‘see here’, and keeping related info close together matters way more now.
Each study here measured something different. DevTune counted citation records, Stack Overflow asked developers, SE Ranking and Ahrefs compared pages and files, and our count is one day’s plain request plus a browser check of 20 pages. None of them compares cited developer-tool pages with similar pages that were not cited, so none can say what causes a citation.
What you can track is the share of your category’s citations landing on pages you can change, before and after each round of fixes.
With a developer tool, we would start with an audit that takes that measurement for you, built from your buyers’ questions and the pages the answers cite. It shows whether the answers cite your docs, your blog or someone else’s listicle, and it becomes a roadmap in quarters, since citations take months to move. A senior GTM consultant of ours owns that roadmap, is on your calls and takes the docs first:
What we would work on
- Docs front pages that return their text to a request without JavaScript, with a Markdown copy where your platform can make one
- Comparison pages and tutorials rewritten where a release, yours or a rival’s, has made them wrong
- Emails to the editors of the developer-tool listicles those answers cite, each making the case for your product
- Replies in the Reddit threads those answers lean on, naming the product only where it answers the question
- A standing report on the share of your category’s citations that your own pages earn
Frequently Asked Questions
Should we rebuild our developer-tool landing page first?
Citation data alone does not justify a rebuild. Landing pages supplied 2.6% of the records in DevTune’s pooled count and were the most-cited own page type on just 19 of 421 vendor domains. Neither figure measures conversions, so judge the landing page by what it does for a developer who already wants to try the product. If your docs or blog fail the readability and freshness checks, fix those first.
Do AI engines read developer documentation?
In this category they cite it often. Documentation was the second-largest page type in DevTune’s data at 17.3% of records, and 21.9% on ChatGPT Search, and it was the leading own page type on 155 of the 421 vendor domains.
Whether an engine can read a particular docs page is a separate question, and no source here measured how each engine treats scripts. What you can test is what the page returns to a request that skips them. Test your docs front page and your ten most-visited docs pages before anything else.
Will publishing an llms.txt get our docs cited?
No study cited here shows that it does. Ahrefs found that 97% of the llms.txt files it watched got no requests at all in May 2026, and SE Ranking found no relation between having the file and being cited. Among developer-tool vendors, 319 of 411 sites already publish one, so it is common in this sample. The file costs little to keep, but the pages it points to have to return readable text first, or it leads engines to near-empty pages.
Does JavaScript-rendered documentation hurt AI visibility?
It can leave a fetcher that skips scripts with little to read. Our count found 61 of 380 docs front pages under 150 words to a plain request. When we opened 20 of them in a browser, 11 of the 18 still short to a plain request passed 150 words once their scripts ran. No source here measures how each engine’s fetcher handles scripts, so test your own pages by comparing view-source with the rendered page, and put the text in the HTML wherever the two differ.
Should we publish our docs as Markdown?
It is a small change with an unmeasured payoff. In our count, 173 of 380 vendors returned a Markdown version of their docs front page when “.md” was added to the address, and the most-cited quarter did so more often, 57.6% against 33.7%. That gap describes who publishes Markdown and proves no effect. If your docs platform can output Markdown from the same source files, ship it. If it cannot, put the text in the server-rendered HTML first.
Which AI engines cite documentation most for developer tools?
In DevTune’s data, ChatGPT Search gave documentation a larger share than any of the other four engines, 21.9%, close to twice Google AI Mode’s 11%. ChatGPT Search also leaned on discussions at 19.3%, while Gemini Search gave blog posts their highest share and Perplexity did the same for best-tools listicles. Ask new signups which assistant they used, or check your referral data, and weight the docs work toward the engines that send you developers.
Do vendor comparison pages get the vendor named in AI answers?
Usually, in DevTune’s data. Across 2,295 pairings of an answer with a ranked page a vendor published, the publisher was named 74.4% of the time. In 17.1% a rival was named while the publisher was left out. On that measure comparison pages reached 78.7% and best-tools listicles 63.2%. A mention says nothing about tone, so state your criteria, treat rivals fairly and check how the answers describe you as well as whether they name you.
Is Reddit worth the effort for a developer tool?
In DevTune’s developer-tool sample, no independent domain drew more citations than Reddit, at 10,831 records across 5,109 URLs. Developers also gather on GitHub and Stack Overflow, but the study gives no comparable count for either. Start with the threads that already show up in answers to your category’s questions. Take part only where the product answers what was asked, from an account with a real history in that community.
Who should own documentation fixes for AI search?
Split the work by the kind of fix. The docs lead or technical writer owns accuracy, structure and whether a page answers a buying question by itself. Engineering owns delivery, meaning server-side rendering, Markdown output, robots rules and the llms.txt file. Marketing owns the set of buyer questions and the engines to test. Name the engineering owner before the work starts, because none of the rendering fixes ships without one.
How do we know whether the work is moving citations for our category?
Take a baseline before changing anything. Choose the questions your buyers ask, run them on the engines they use and log every cited page by type and owner. Rerun the same questions on a fixed schedule and watch the share of citations on pages you control, docs included. Read that share as an early signal and AI-referred signups as the outcome, kept in separate columns so one never stands in for the other.
Sources (9)
- DevTune, "The State of AI Search for Dev Tools" (2026)↩1↩2↩3↩4↩5↩6↩7↩8↩9↩10↩11↩12↩13↩14↩15↩16↩17↩18↩19↩20↩21↩22↩23↩24↩251 July 2026. 43 developer-tool categories, 478 tracked brands, 1,075 test questions re-run weekly on five AI engines, 126,814 citation records, 2 April to 30 June 2026. The vendor-level counts are our arithmetic on its published brand table. Read 13 November 2026.
- Stack Overflow, "2026 Developer Survey" (2026)↩1↩2↩3↩4↩5↩6↩7↩8↩930,903 responses from 169 countries, 23 June to 5 August 2026; each question has its own base. Read 12 November 2026.
- Empact Partners, "vendor-site count" (2026)Empact Partners' own data↩1↩2↩3↩4↩5↩6↩7↩8the 421 domains of the brands in DevTune’s 2026 benchmark, each requested once without JavaScript on 13 October 2026; 411 answered. Most-cited and least-cited quarters by DevTune’s citation records, 102 sites each. Browser check of 20 near-empty documentation front pages, with a same-day plain request, 13 November 2026.
- Ahrefs (Glen Allsopp), "Do Self-Promotional ‘Best’ Lists Boost ChatGPT Visibility? Study of 26,283 Source URLs" (2025)↩1↩2↩34 December 2025. 750 ChatGPT prompts across software, products and services. Read 12 November 2026.
- Ahrefs, "We Analyzed 137K Sites: 97% of llms.txt Files Never Get Read" (2026)↩115 June 2026. 137,210 domains using Ahrefs Web Analytics, May 2026. Read 12 November 2026.
- SE Ranking, "Does LLMs.txt impact your AI visibility and citations? No, according to research" (2025)↩17 November 2025. Nearly 300,000 domains. Read 12 November 2026.
- SE Ranking, "How to optimize for ChatGPT: LLMs.txt doesn’t matter but brand mentions on Quora and Reddit do" (2025)↩124 November 2025. 129,000 domains and 216,524 pages across 20 niches. Read 12 November 2026.
- Ahrefs, "New Study: AI Assistants Prefer to Cite ‘Fresher’ Content (17 Million Citations Analyzed)" (2025)↩128 July 2025. Read 12 November 2026.
- Similarweb (Maayan Zohar Basteker), "The Most Cited Domains by LLMs and the Signals That Get You Cited" (2026)↩114 April 2026. Nearly 600,000 citation events, United States, January to February 2026. Read 12 November 2026.
Keep reading
Statistics 43 SaaS Community Statistics
43 figures on what SaaS communities cost and return, and which ones AI answers cite: Reddit far more than any vendor’s own forum.
Statistics 16 AI Citation Concentration Statistics for SaaS
AI citation concentration statistics to help SaaS teams choose which cited pages to pursue and compare engines without mixing measures.
Statistics 48 AI Search Citation Statistics
See which pages AI answers cite, why your own site is only part of the picture, and where to focus your SaaS visibility budget.


