Back to BlogAI & Automation

We Ran 1,429 AI Searches. The Engines Disagree.

Ryan Goering
5 min read
We Ran 1,429 AI Searches. The Engines Disagree.
Free Agentic Browsing Audit

Can an AI agent actually use your website?

AI agents are starting to visit sites on a customer's behalf — comparing companies and requesting quotes. We run the same agentic browsing checks Google built into Chrome and show you what an agent hits when it lands on your site. No email to run it.

Last updated: September 2026.

By Ryan Goering, founder and CEO of BaaDigi, a contractor marketing agency.

Study period: 18–31 August 2026. Methodology and limitations in full below.

The headline table

DomainTotalChatGPTPerplexityGoogle AI Overview
reddit.com657136382139
baadigi.com (us — see disclosure)38212119962
google.com31000310
linkedin.com291624639
youtube.com289471214
bluecorona.com14168451
thriveagency.com1321510512
agencies.semrush.com127151039
hookagency.com12071067
sianamarketing.com9434519
beseencontractors.com8646400
searchengineland.com8128476

The table lists the domains discussed in this article. Three domains fell inside the same range and are omitted only for length: pipelineon.com (91 citations: 16 ChatGPT, 61 Perplexity, 14 AI Overview), firstpagesage.com (90: 15, 57, 18) and webfx.com (84: 13, 62, 9). Every count was re-checked against the source table on 22 September 2026.

Finding 1: the engines are not interchangeable

The clearest result. Two domains show near-total divergence:

  • LinkedIn — 246 Perplexity citations, 6 in ChatGPT. A 41× difference.
  • YouTube — 214 AI Overview citations, 4 in ChatGPT. A 53× difference.

Advice to "publish on LinkedIn for AI visibility" is effectively Perplexity-specific advice. Advice to "make videos for AI search" is effectively Google AI Overview advice. Neither is wrong, but presenting either as general AI-search strategy is misleading, and most published guidance does exactly that.

Finding 2: forums and social outrank agencies

Reddit was the single most-cited domain at 657, ahead of every agency and publisher in the set. LinkedIn and YouTube also outranked every marketing agency measured.

On buyer-evaluation questions especially — is this platform worth it, what does it really cost — the engines reach for peer discussion over vendor pages. That is rational: a contractor describing what they actually paid is better evidence than an agency describing what contractors typically pay. It also means an agency competing on "we wrote a thorough page" is competing against thousands of people with direct experience.

Is your own website costing you leads?

Free scan — speed, mobile, local visibility, and lead capture. Results in about 30 seconds.

Finding 3: Google cites itself, heavily

google.com appeared 310 times, exclusively in AI Overviews and never in ChatGPT or Perplexity. Google's own properties — support documentation, Business Profile help, Ads guidance — are a substantial share of what its AI surfaces cite. For any question where Google has first-party documentation, that documentation is a strong default citation you are competing against rather than displacing.

Finding 4: some sources are engine specialists

beseencontractors.com earned 46 ChatGPT citations and zero in AI Overviews. Conversely bluecorona.com skewed toward Perplexity (84) and AI Overviews (51) with only 6 in ChatGPT.

Nobody appears to be winning all three. Which suggests the engines weight different signals rather than converging on one quality standard — and that measuring a single engine will mislead you about your actual position.

What we would do with this

  1. Measure per engine, never blended. A blended citation rate hides the case where you are strong in one engine and absent in another.
  2. Decide which engine matters for your buyers rather than chasing all three. They have very different user bases and volumes.
  3. Take peer platforms seriously. Reddit, LinkedIn and YouTube outrank every agency site here. Being genuinely present and useful there is not a content-marketing tactic; it is where the engines are actually reading.
  4. Do not trust single-engine case studies, including the ones agencies publish. A result in Perplexity may say nothing about ChatGPT.

Methodology

Sample. 1,429 answers collected 18–31 August 2026 across 88 prompts, run Monday/Wednesday/Friday. Engine split: Perplexity 511, ChatGPT 511, Google AI Overview 407. Citations are the source URLs each engine returned with its answer; we counted domains, not unique pages.

Prompt selection is not neutral, and this matters. The 88 prompts are contractor-marketing questions we chose — agency selection, lead-platform evaluation, cost questions, local SEO. They are not a random sample of what people ask AI, and results would differ for another topic. A study of cooking or legal questions would produce a different table.

We are in our own data. baadigi.com is the second most-cited domain here. That is partly because several prompts name us or our city, which guarantees a citation and measures nothing. Read our row as an artefact of the method rather than a finding, and read the third-party rows as the useful part.

Other limitations. Fourteen days is a short window. AI answers vary between identical runs, which is why repeated sampling matters. Google AI Overview coverage requires asynchronous loading; runs where no AI Overview was served are excluded rather than counted as zero. Engine behaviour changes without notice, so treat these as a snapshot, not a constant.

Corrections welcome. If you run comparable tracking and get different results, we would genuinely like to see it.

Related

How to get cited by AI · How to show up in AI search · Check your own AI visibility

Related: AEO vs SEO for contractors · Which AI engine actually sends contractor leads? · LLM SEO (definition)

Free website audit — see what's costing you leads

No email required. Takes 2 minutes. Completely free.

ai searchstudyoriginal researchchatgptperplexityai overviews
Ryan Goering

Ryan Goering

CEO & Founder, BaaDigi

U.S. military veteran and digital marketing strategist who built BaaDigi to help contractors generate predictable leads and revenue. 15+ years in SEO, PPC, and AI-powered marketing automation.

Frequently Asked Questions

Do ChatGPT, Perplexity and Google AI Overviews cite the same sources?

No. Across 1,429 tracked answers to 88 contractor-marketing questions, the three engines cited almost entirely different domains. LinkedIn appeared 246 times in Perplexity and only 6 times in ChatGPT. YouTube appeared 214 times in Google AI Overviews and 4 times in ChatGPT. Optimising for one engine says little about your position in another.

Which sources do AI search engines cite most for contractor marketing questions?

In this sample Reddit led with 657 citations, ahead of every marketing agency and publisher measured. LinkedIn and YouTube also outranked every agency site. On buyer-evaluation questions the engines favour peer discussion over vendor pages, which is rational since a contractor describing what they actually paid is stronger evidence than an agency describing typical costs.

Does Google cite its own properties in AI Overviews?

Heavily. google.com appeared 310 times in this sample, exclusively within Google AI Overviews and never in ChatGPT or Perplexity. For questions where Google holds first-party documentation such as Business Profile or Ads help, that documentation is a strong default citation competing for the same slot.

Can a website get cited by all three AI engines?

None of the domains in this sample managed it evenly. beseencontractors.com earned 46 ChatGPT citations and zero in AI Overviews, while bluecorona.com skewed to Perplexity and AI Overviews with only 6 in ChatGPT. The engines appear to weight different signals rather than converging on a single quality standard.

How was this AI citation study conducted?

1,429 answers were collected between 18 and 31 August 2026 across 88 prompts run three times weekly, split as Perplexity 511, ChatGPT 511 and Google AI Overview 407, counting cited domains rather than unique pages. Prompt selection was not neutral — they are contractor-marketing questions chosen by us — and the publisher appears in its own data, which is disclosed in the methodology.

Ready to Get More Leads?

Get a free marketing audit for your contracting business. We'll show you exactly where you're losing leads.

Get Your Free Audit