AI Citation Map: see the pages AI reads before it recommends software

How to use the AI Citation Map

Seven steps, about two minutes. The map is the same for everyone in a category; your website only changes which rows are marked as yours.

1. 1

   Pick your software category

   Choose the category your buyers would type into ChatGPT, such as CRM or E-signature. The map covers 20 B2B categories. If yours is close but not exact, pick the nearest one: many of the cited sites overlap between neighbouring categories.

   Software category

   CRM

   AccountingCRME-signatureEmail marketing
2. 2

   Enter your website and brand name

   Add your domain so we can mark your own rows and check the list pages for you. The brand name is optional but worth adding if it differs from your domain, because AI answers name brands, not URLs.

   Your website

   yourbrand.com

   Brand name

   YourBrand

   Show my citation map
3. 3

   Read the four numbers at the top

   AI answers analysed is the size of the sample for this category. Different sites cited shows how spread out the sources are. Answers citing your site is how often an engine linked to you as a source. Answers naming your brand is how often you were mentioned at all. A brand can be named without being cited, which means AI is learning about you from other people’s pages.

   48AI answers analysed

   148different sites cited

   0%answers citing your site

   6%answers naming your brand

   Illustrative values for a brand not yet cited.

4. 4

   Switch between the engine tabs

   The engines behave very differently, so check each one. In our data ChatGPT cites far more sources per answer than the others, while Google AI Overviews lean heavily on YouTube and Reddit. A gap on one engine usually points to one type of source you are missing.

   All enginesChatGPTGeminiAI ModeAI Overviews

   Every number on the result recalculates for the engine you pick.

5. 5

   Read the most-cited sites and the source mix

   The table ranks sites by the share of answers that cite them, with dots for the engines that used each one. The source mix shows which kinds of sites AI trusts in your category: review sites, publishers, forums, video, blogs or the vendors themselves. That tells you where to spend effort first.

   1. 1salesforce.com40%
   2. 2hubspot.com40%
   3. 3reddit.com38%
   4. 4zapier.com35%

   Real CRM data: share of answers citing each site.

6. 6

   Work through the pages to get onto

   These are the exact pages AI cited most. Each one carries a status. Lists you: the page already mentions you, so keep it accurate and up to date. Not listed: a list or review page that skips you; this is your outreach list. Join the thread: a forum discussion where the way in is a genuine, helpful reply. Video: a video AI cited; pitch the creator or make your own. List page or Not checked: we could not confirm either way, so open the page and look.

   Best CRM software for small businessespublisher.com

   Not listed

   15 CRM tools comparedblog-site.com

   Lists you

   Which CRM do you use and why?reddit.com

   Join the thread

   Illustrative rows. Your result lists the real pages.

7. 7

   Share or save the link

   The address bar updates with your category and website, so the result can be shared with your team or reopened later. Use the Copy link button at the bottom of your result.

   topsearchagency.com/ai-citation-map/?cat=crm&domain=yourbrand.com

   Copy link

What an AI citation is, and why it decides which software gets recommended

When someone asks ChatGPT, Gemini or Google’s AI features a question with web search switched on, the engine reads a handful of web pages and writes its answer from them. The pages it links to are its citations. They are the closest thing we have to seeing what the AI read before it answered.

For software buyers this matters more than it first appears. Ask “what is the best CRM for a small business?” and the brands in the answer are mostly the brands that appear on the pages the engine cited: a publisher’s round-up, a review site, a Reddit thread, a vendor’s comparison page. If your product is not on those pages, it rarely makes the answer, however well your own site ranks on Google.

Cited versus mentioned

The map tracks two different things. A citation is a link to your website as a source. A mention is your brand name appearing in the answer text. Many brands are mentioned without being cited, which means the AI formed its view of them from other people’s pages. Those third-party pages decide how you come across, which is why this map focuses on them.

Why a small set of pages carries so much weight

Across 949 answers, the engines cited 1,800 different websites. That sounds spread out, but the same few sites keep returning: three of them, Zapier’s blog, Reddit and YouTube, appeared in the answers of all 20 categories we tested. Getting onto the right handful of pages in your category is a realistic project, not an endless one.

Key findings from 949 AI answers

We asked 12 buyer questions in each of 20 B2B software categories on four AI engines, then logged every source cited. These are the patterns that held across the whole sample.

1. A few sites show up everywhere

zapier.com was cited in 34% of answers, reddit.com in 22% and youtube.com in 21%. All three appeared in every one of the 20 categories. Zapier’s lead comes from its large library of “best apps” articles, which engines treat as neutral round-ups. For most software brands, these three are the first places to check.

2. Independent blogs and list sites carry the most weight

By number of citations, blogs and list sites were the largest group, followed by the vendors’ own websites and then publishers. Forums, review sites and video each took a smaller share of citations, but they appear in a large share of answers: forum or community pages in 31% of answers and video in 21%.

3. Your own website still matters

52% of answers cited at least one vendor’s own site, usually a product, pricing or comparison page. Engines use vendor pages to confirm facts, so clear, specific pages about what you do, who it is for and what it costs give them something to cite.

How ChatGPT, Gemini, AI Mode and AI Overviews differ

The four engines answered the same questions, yet they leaned on very different sources. Winning on one does not mean you are visible on the others.

ChatGPT: many sources, almost no Reddit or YouTube

ChatGPT cited an average of 14.6 sites per answer and 1,198 different sites in total, far more than any other engine. It cited a vendor’s website in 98% of answers and a review site in 35%, but Reddit in only 0% and YouTube in 0%. For ChatGPT, list articles on independent sites and your own product pages do most of the work.

Gemini: few sources, mostly publishers and blogs

Gemini cited 3.2 sites per answer on average. zapier.com led with 29% of answers, and Reddit appeared in 14%. With so few sources per answer, each cited page has a large influence on what Gemini says.

Google AI Mode: Reddit and big publishers

AI Mode cited the fewest sources, 2.9 per answer, and leaned on Reddit (29% of answers), YouTube (19%) and large publishers such as pcmag.com.

Google AI Overviews: YouTube first

AI Overviews, the summaries at the top of Google search, cited YouTube in 67% of answers and Reddit in 44%, more than any other engine. If buyers in your category search Google, a useful video and an honest presence in the relevant Reddit threads count for a lot here.

Category highlights

Each category has its own set of trusted sources. Use the tool above for the full map of yours; here is the single most-cited site in each.

Where one site dominates

In Scheduling & booking, zapier.com was cited in 70% of answers, and in Team chat in 62%. In categories like these, one well-placed mention can reach most of the AI answers buyers see.

Where small independent sites lead

In AI meeting assistants the most-cited source was a single Medium post (justtalkingtech.medium.com, 33% of answers), and in Applicant tracking (ATS) it was theundercoverrecruiter.com (35%). Newer categories have fewer established review pages, so a focused article from a smaller site can shape the answer.

Where vendors cite each other

In CRM, salesforce.com and hubspot.com were each cited in 40% of answers, mostly through their own guides and comparison pages. In Password managers, bitwarden.com led with 36%. Market leaders publish content that engines cite, which is a lesson for challengers too.

The roles of Reddit, YouTube and review sites

Reddit: the voice of real users

Reddit was cited in 44% of Google AI Overviews and 29% of AI Mode answers, but in only 0% of ChatGPT answers. Engines use threads to check what practitioners actually recommend. You cannot buy your way into these threads, and trying usually backfires. What works is staff and happy customers answering honestly where your product fits, and being specific about who it is not for.

YouTube: Google’s favourite source

YouTube was cited in 67% of AI Overviews, the highest rate of any site on any engine. Reviews, walkthroughs and comparison videos are what get cited. Two routes: help creators who already review your category, or publish your own clear product walkthroughs with descriptive titles and chapters.

Review sites: proof engines can check

Review sites such as G2, Capterra and Gartner were cited in 18% of all answers and 35% of ChatGPT answers. G2 alone was cited in 7% of answers across 18 categories. A complete profile with recent reviews is a low-effort way to appear in these citations.

How to get onto the pages AI cites

Six steps, in the order that usually pays off fastest. Use your map’s “Pages to get onto” list as the to-do list.

1. 1

   Start with the list pages marked Not listed

   Open each one and find who maintains it. Many round-ups are updated every few months and the editor accepts suggestions. Send a short, specific note: what your product does, which reader it suits, one fact they can verify, and a link to a free trial or demo account. Do not offer payment for inclusion on editorial lists.

2. 2

   Complete and refresh your review-site profiles

   Fill in every field on G2, Capterra and any review site in your map, then ask recent happy customers for reviews. Recent, detailed reviews give engines specific, current statements to draw on.

3. 3

   Join the forum threads that keep getting cited

   For each thread marked Join the thread, read it fully and reply only where you add something useful. Say who you are. Answer the question even if the honest answer is a competitor for some buyers. One good reply in a cited thread is worth more than dozens in threads nobody cites.

4. 4

   Get into the videos AI cites

   Look at the cited videos in your map and contact those creators with a demo account and a clear angle. In parallel, publish your own walkthrough and comparison videos with plain titles that match the buyer questions.

5. 5

   Make your own site worth citing

   Engines cite vendor pages to confirm facts. Publish pages that answer buyer questions directly: pricing, who the product is for, comparisons with the main alternatives, and integrations. Keep them factual and up to date; outdated pricing gets you described wrongly.

6. 6

   Re-check every month

   AI answers change. Re-run the map for your category monthly, track which pages now list you, and move to the next ones on the list.

Methodology and limits

How the data was collected

• Questions: 12 buyer questions per category, the same pattern in every category: the best tool overall, for small businesses, for startups and for large companies; the most affordable; the easiest to use; what experts recommend; alternatives to the two market leaders; a head-to-head between them; a comparison; and whether a free option is any good.
• Engines: ChatGPT and Gemini with web search, Google AI Mode, and Google search with AI Overviews. Each question was asked once on each engine.
• Location and context: United States, English, no logged-in history.
• Sample: 949 usable answers (239 ChatGPT, 240 Gemini, 240 AI Mode, 230 AI Overviews). AI Overviews only count where Google showed one.
• Date: collected 2026-11-15. The date on each result shows when its data was gathered.
• Citations: every linked source was cleaned (tracking parameters and fragments removed) and counted once per answer per site. Image and logo hosts were excluded.
• Page checks: the most-cited list, review and publisher pages in each category were visited once to record which brands they mention and which sites they link to. Some sites block automated visits; those show as Not checked.

Limits to keep in mind

• AI answers vary between runs, accounts and locations. Treat the map as a strong signal of which sources matter, not a fixed ranking.
• About 48 answers per category is enough to see clear patterns, but small differences of a few percentage points between sites are not meaningful.
• Site types are assigned automatically, so the odd site may be labelled imperfectly.
• Brand mentions are matched by name, so very generic brand names can be over- or under-counted.

Questions about the AI Citation Map