8 of 12 Top AI Answer Sources Aren't Competitors

Q: What did GrowByData claim?

GrowByData reported that across 4,011 captured AI answers covering 44 search and AI-visibility prompts, 8 of the 12 most frequently appearing domains were outside the tracked competitor set. YouTube, Reddit, Search Engine Land, and LinkedIn each appeared in more than 10% of observations.

Q: Is this independent research?

No. GrowByData sells AI visibility tracking products, so the findings should be treated as vendor research. The study can be useful as a directional signal about source mix, but teams should replicate the analysis using their own prompts, competitors, and target category.

Q: Does this replace watching SEO competitors?

No. Traditional SEO competitors still matter for rankings, traffic, and sales comparisons. AI search adds a second competitive layer that includes publishers, directories, community platforms, and other domains that repeatedly appear inside AI-generated answers. Teams should track both sets separately.

Q: Why do retail and software look different?

Source mix varies by category. In GrowByData's software dataset, vendor domains appeared alongside publishers and platforms. In its retail spot check, publishers and review sites dominated the top results with no product-brand sites in the top 10. This suggests that AI source composition should be measured by category rather than assumed to be universal.

Q: How should reporting change this week?

Separate reporting into owned-domain presence, tracked competitor presence, and third-party presence such as publishers, user-generated content platforms, and directories. Report those metrics by engine rather than combining them into a single AI visibility score, which can hide source gaps on platforms such as YouTube, Reddit, or industry roundups.

Q: How does this relate to citation overlap?

Citation overlap studies show that different AI engines often cite different URLs for the same prompts. GrowByData's study adds another dimension by showing that many frequently appearing domains are not the competitors already tracked in traditional SEO tools. Together, the findings support monitoring both engine-specific citation behavior and a broader set of third-party source domains.

Written byPawan BhargavSr. Content Writer, DerivateX

Reviewed byFouziaSEO Growth & Operations, DerivateX

Published Sep 30, 2026Updated Sep 29, 2026

6 min read

TL;DR

On September 24, 2026, GrowByData published a study of 44 prompts across ChatGPT, Google AI Mode, Google AI Overviews, and Perplexity, capturing 4,011 answer observations (Aug 4 to Sep 2, 2026, US/NY).
Among 31 domains reviewed, 8 of the 12 most-present were outside the configured competitor set. YouTube appeared in 799 observations (19.9%), Reddit in 694 (17.3%), Search Engine Land in 559 (13.9%), and LinkedIn in 412 (10.3%).
The most-present untracked vendor hit 473 observations (11.8%), more often than 12 of 15 vendors inside the tracked set. A two-prompt retail spot check put only publishers and review sites in both top 10s; no product brand site made either list.
Pair this with DerivateX’s directories on supplier prompts and ~9% citation overlap: the SEO competitor list, the directory layer, and the cross-engine source mix are three different maps.
GrowByData sells LLM Intelligence / Compass tracking, so treat the series as a vendor study with a product stake, then rebuild the source list on your own buyer prompts. Start with a free AI visibility audit if you need owned vs third-party presence split before you reallocate PR and content.

Your rank tracker answers who fights you for blue links. It does not answer who is feeding the answer your buyer reads before they ever click.

GrowByData’s September 24 study puts a number on that gap for a software-and-search prompt set: most of the domains that kept showing up were not the vendors the tracker was built to watch. For B2B SaaS GEO, that is the difference between “we monitor our five rivals” and “we monitor the publishers, communities, and untracked tools that actually shape shortlists.”

What GrowByData measured

From the primary write-up by Manogya Guragai:

Scope: 44 prompts about search and AI visibility, four platforms, New York location, Aug 4 to Sep 2, 2026. An observation = one prompt answered on one platform on one day. Base for every share = the 4,011 captured observations (not the theoretical 5,280).
Presence definition: a domain counted if it was cited or linked anywhere in that answer. They reviewed 15 tracked vendors plus 16 other domains that appeared.
Top of the presence table (selected): Tracked vendor 1 at 24.7%; YouTube 19.9%; Tracked vendor 2 19.8%; Reddit 17.3%; Tracked vendor 3 16.7%; Search Engine Land 13.9%; Untracked vendor A 11.8%; LinkedIn 10.3%.
Retail spot check (two prompts, 46 observations each): mattress-in-a-box and overnight acne stickers. Leading sources were Sleep Foundation, Forbes, Mattress Nerd, RTINGS, NBC News, Allure, NYT, Healthline, and similar. Zero product-brand sites in either top 10.

GrowByData’s own framing: “Your SEO competitor list and your AI-answer source list are not the same thing.”

What the study measures vs what it misses

Useful for operators

Hard evidence that platforms (YouTube, Reddit, LinkedIn) and trade pubs sit inside the AI source mix beside vendors.
Proof that an “untracked” software vendor can out-appear most of a carefully configured competitor set.
A clean reminder that retail and software categories do not share the same source mix (vendors still show in software; publishers dominate the retail spot check).

Caveats (do not skip)

GrowByData sells Compass / LLM Intelligence. Vendor research has a stake in making multi-source tracking feel urgent. Replicate on your category before a board claim.
The 44 prompts are one configured set (mostly informational). They are not a census of every B2B buyer question.
Shares combine ChatGPT, AI Mode, AI Overviews, and Perplexity. Platform-level mixes can diverge; the study itself warns against assuming one source pie.
Presence is not recommendation strength and is not referral traffic. A Reddit thread can shape the answer without sending a session to your site.
Missing observations (5,280 theoretical vs 4,011 captured) were not attributed. AI Overviews appeared less often than ChatGPT or AI Mode in this set.

How this fits DerivateX’s existing maps

Map	What it answers	What GrowByData adds
Directories on supplier prompts	How much third-party directory / listicle weight shows up on commercial supplier asks	Even outside “directory” labels, UGC and trade pubs sit in the top presence tier
~9% citation overlap	Engines do not reuse the same URLs	Inside one blended view, the types of domains still expand past your competitor set
Citation Surface Map	Where corroboration lives (owned, influenced, earned)	Your monitoring list must include platforms and pubs that never show up in keyword-overlap tools

Weak read: “Stop watching competitors; only do Reddit and YouTube.”

Better read: Keep the competitor set for sales and SERPs. Add a second list for AI source presence (publishers, communities, review sites, directories, and vendors your keyword tool never named).

Operator playbook / worked example

Layer 1: Build two lists, not one

SERP competitor list: domains that fight you for rankings and paid.
AI source list: every domain that appears repeatedly across your buyer prompts for 4+ weeks, typed as owned / vendor / publisher / community / review / directory.

If list 2 is mostly a copy of list 1, your tracking is incomplete, not “clean.”

Layer 2: Score presence by source type weekly

For a fixed prompt frame (ChatGPT, Gemini/Claude if relevant, Perplexity, AI Overviews / AI Mode):

% of answers with your URLs
% with named competitors
% with publishers / UGC / directories that mention the category without you

Budget follows the gap. Missing owned pages means extractable first-party answers. Missing you on roundups that already feed answers means editorial and digital PR, not another blog post that restates the category.

Layer 3: Worked example (mid-market RevOps SaaS)

Priority prompts: “best RevOps platform for Series B,” “HubSpot RevOps alternatives with Salesforce sync,” “pipeline hygiene software for B2B.”

Export who your SEO tool calls competitors (usual five to eight).
Run the three prompts weekly across four engines. Log every cited domain for 30 days.
Tag each domain: tracked rival, untracked vendor, trade pub, Reddit/YouTube/LinkedIn, review/directory, other.
If Search Engine Land, G2, or a Reddit thread appears more often than rival #4, put those URLs on the weekly standup. Fix what they say about you (or the silence).
Separately track whether your comparison and pricing pages ever appear. GrowByData’s retail spot check is a warning: category answers can be built entirely from third parties.

Layer 4: Consistency still matters

GrowByData measures who shows up. Pair it with retention thinking from studies like Techmagnate’s LLM Citation Drift Report (India personal-loan category, Sep 22, 2026: about 39% average weekly domain churn; about 15% of domains held every week and captured about 93% of citations). Different vertical, same operator lesson: a one-week mention is not a durable shortlist seat. Techmagnate sells Prism / GEO services, so treat that series as a second vendor panel and verify on your prompts.

How DerivateX thinks about it

Rank trackers and AI source monitors answer different questions. Do not let one dashboard pretend to be both.
Treat GrowByData as a source-mix signal with a measurement-product COI, then rebuild the presence table on your ICP prompts.
Separate owned extractability, influenced third parties, and earned community/platform work. YouTube and Reddit are not “content ideas”; they are surfaces already inside answers.
Keep engine-level tracking. A blended 4,011-observation pie is a diagnosis, not a single KPI to optimize.

If you want a baseline that splits owned vs publisher vs community presence on your category, request a free AI visibility audit. Engagement options stay on pricing. Adjacent reading: directories on supplier prompts, citation overlap, the Citation Surface Map, and case studies where third-party corroboration showed up in CRM.

FAQ

What did GrowByData claim?

That across 4,011 captured AI answers on 44 search/AI-visibility prompts, 8 of the 12 most-present domains were outside the tracked competitor set, with YouTube, Reddit, Search Engine Land, and LinkedIn each above 10% of observations.

Is this independent research?

No. GrowByData sells AI visibility tracking products. Use the study as a directional source-mix signal and replicate with your own prompts and competitor set.

Does this replace watching SEO competitors?

No. Keep SERP competitors for rankings and sales. Add a second list for domains that repeatedly appear inside AI answers, including platforms and publishers.

Why do retail and software look different?

In GrowByData’s software set, vendors still appear alongside platforms and pubs. In the two-prompt retail spot check, publishers and review sites filled both top 10s with no product brand sites. Category source mix is not universal.

How should reporting change this week?

Report owned presence, tracked-rival presence, and third-party (publisher/UGC/directory) presence separately, by engine. Stop one blended “AI visibility” score from hiding a YouTube or roundup problem.

How does this relate to citation overlap?

Overlap studies show engines reuse few of the same URLs. This study shows that even inside a blended view, many of the URLs that do appear are not the rivals your SEO tool already monitors.