Somewhere right now, a guy with 4,000 YouTube subscribers and a Camry in his driveway is making your next customer's purchase decision for them.
He doesn't run an aftermarket ecommerce store – he's not trying to.
He just made a decent video about a flickering headlight relay two years ago, explaining the fix, recommending a part, and linking to Amazon.
Google's AI Overview shows his video to tens of thousands of people troubleshooting that issue. Your sale is gone before any parts e-tailer ever entered the picture.
This is not a rare edge case. It is what our data shows happening at scale across one of the most searched-for parts categories in aftermarket retail.
We analyzed 27,094 headlight-related queries, pulled live citation data on 114 of the messiest diagnostic queries in the category, and analysed searches through Claude and ChatGPT to understand how AI assistants handle the same questions.
The findings apply well beyond headlights. They describe a structural shift in how aftermarket parts customers find information and make purchase decisions – and what you can do about it.
Why we chose headlights
As a product category, headlights span nearly every type of search query that exists in aftermarket retail: customers search to diagnose symptoms, compare products, find fitment-specific parts, research brands, and locate local installation.
It's also a segment that includes both replacement parts and upgrades and modifications.
That makes it a great lens for studying how AI search behaves across a whole market, not just one type of consumer or query. Whatever is true here is most likely also directionally true in every other aftermarket parts category.
AI search has split your market into two separate games
Across the full 27,094-keyword dataset, Google's AI Overview appeared on 41.9% of queries overall. But within that average, there's significant variation in how often AIO shows up, based on search intent:
When a user searches to compare two products, or to diagnose a problem with their vehicle, Google answers the question directly on the results page the vast majority of the time. The user gets what they need without clicking through to anyone's website.
But when a user searches for a specific part for a specific vehicle, or for a specific brand, conventional search results still dominate. The click still happens. The store that ranks well still gets the visit.
Three independent signals in our data all point to the same split.
Searches containing comparison language (like "vs"), question words, or qualifiers like "best" or "brightest" trigger AI Overview 92–97% of the time, while searches containing "buy" trigger it only 15–24% of the time.
Advertiser behavior confirms it: queries with lower commercial intent and lower competitive density are significantly more likely to show AIO, while the queries advertisers fight hardest over — transactional, branded, fitment — are the least likely to show AIO.
And our targeted probe of 114 diagnostic queries confirmed that even when the answer is an ambiguous, "it depends", AIOs trigger 84.2% of the time.
The practical implication: a meaningful share of your potential customers are now having their questions answered by Google before they reach any retailer's website.
This is not a problem that better product pages or stronger rankings will solve for that slice of demand.
But for fitment, branded, and transactional queries, conventional SEO and paid search still function largely as they always have. Those searches still send clicks. That half of the market deserves continued conventional investment.
A balanced strategy addresses both types of searches, and each one requires fundamentally different approaches.
Who is winning AIO citations, and in what format?
For aftermarket parts queries where AIO is the dominant result type, who gets cited most often?
We took our 114 diagnostic queries and pulled the full AI Overview content for each, including every cited source. 96 of the 114 returned usable AI Overview responses, generating 817 total citations — an average of 8.5 citations per query.
Parts retailers and manufacturers — the businesses whose livelihoods depend on selling these parts — collectively account for 11.1% of citation real estate. YouTube alone takes more than three times that share.
Clearly, YouTube content wins AIO on diagnostic queries.
We identified the channel behind each of the 102 unique cited videos and classified each by type:
Nearly three-quarters of the videos winning AI citations in this category come from small, independent creators.
No single channel dominates, with most appearing only once or twice across the citations in our dataset.
Which is not to say that Google prefers to cite independent creators over established brands. To the contrary, where content from established brands is availble, Google does cite them, as shown by the 13.4% of cited videos from recognizable names like 1A Auto and O'Reilly.
The opportunity exists; few parts e-tailers have tapped into it yet.
Also, the median age of a cited video is 3.3 years, with a quarter of cited videos between 5 and 10 years old. Well-targeted diagnostic video content holds its citation value for years.
New entrants are not locked out either. About 20% of cited videos are less than a year old.
Where are the cited videos linking to?
Of the 102 unique cited videos, only 59 videos contained product links. Of these, 71.2% linked to Amazon.
The next largest destination is a single auto-advice content site at 6.9%. Every specialty aftermarket retailer in the dataset combined accounts for under 4%.
Played out in full:
- A user searches for a diagnosis
- AI Overview answers the question and cites a YouTube video
- The user watches the video, trusts it, and clicks the product link in the description
- That link goes to Amazon
Simply because Amazon Associates is the default affiliate program for independent creators, and nobody in aftermarket retail has made it worth their while to link elsewhere.
More on this in the recommendations section.
Who is getting cited in ChatGPT and Claude?
As of June 2026, ChatGPT claims over 1 billion monthly active users, while Claude has an estimated 30 million.
Those numbers are already big enough that you should pay attention, but what's more, data suggest that users referred from LLMs are 23x more likely to convert than those from traditional search.
To get a sense of which parts stores are getting recommended by LLMs, we tested 168 controlled queries on both Claude and ChatGPT, with two phrasings each: search phrases and natural-language-style. Here's what we found.
The gap might have to do with how often each platform searched the live web before answering: Claude triggered a web search on 63.1% of queries overall; ChatGPT on 33.3%.
An LLM that retrieves current sources tends to name real, specific entities. One that answers from training data alone tends toward generic advice with no named brands.
On symptom-diagnostic queries specifically — the query type most likely to precede a parts purchase — Claude searched the web 91.7% of the time. ChatGPT searched 0% of the time on those same queries, answering entirely from training data with no real-time grounding and no named sources.
On branded and near-me/install queries, both platforms searched at similar rates. The gap is concentrated exactly where it matters most for parts retail.
Caveat: This was measured via API, so the exact percentages may differ from behavior in the consumer-facing app. Also, LLM behavior changes frequently. We'll update this section as we gather newer data.
Recommendations for parts ecommerce stores
The data above points to three distinct actions.
Not all of them are equally important for every store, but they address different parts of the organic revenue equation.
1. Don't drop the ball on "traditional" search
The noise around AI search can make it sound like conventional SEO and paid search are becoming obsolete.
For ecommerce at least, they are not.
Fitment queries (no AIO on 65% of searches), branded queries (no AIO on 72.7% of searches), and transactional queries (no AIO on 53.2% of searches) still resolve as a normal click-based results page the majority of the time.
Product content, technical fitment data, category page optimization, and paid search on high-intent terms still work the same way they always have.
Do not deprioritize this investment in response to AIO.
2. Build diagnostic video content under your own channel
For the diagnostic and comparison queries where AIO dominates, the format that wins citation is YouTube video — cited in 93% of usable AI Overview responses in our data.
The competition is not what you might expect: nearly three-quarters of cited videos come from small independent creators.
The bar to meet is specificity and usefulness, not production quality.
A video titled "headlight flickering fix" is generic; but a video about "2018 Honda Accord headlight flicker fix – 5 different solutions," is specific. On relevance alone, the second one is far more likely to get cited to that specific search intent.
Diagnostic video is also a durable asset. The median cited video in our dataset is 3.3 years old.
A focused library of well-targeted troubleshooting content is an enduring investment.
Check out tools like KeyShot and Cortona to help reduce the need for expensive, time-consuming shoots.
3. Reach out to the creators already winning citations, and offer them a better deal than Amazon
Independent creators already earning citations in your subcategory are linking their viewers to Amazon by default, not out of loyalty.
An affiliate arrangement with meaningfully better terms, a direct relationship, and possibly free product to feature changes that calculus.
You are not asking these creators to change or create new content. You are just asking them to swap one link for another, for more money, in a description box they have already written.
Finding five to ten relevant creators and converting even a few of them redirects existing citation traffic that is already happening, without waiting to build a video library from scratch.
Methodology: findings are drawn from a 27,094-keyword cleaned SEMrush export covering the U.S. headlights category, a 114-queryhand-built probe of ambiguous diagnostic searches verified via live SERP API data (DataForSEO), citation analysis across 817 individual citations from 96 usable AI Overview responses, a 102-video YouTube citation deep-dive with outbound link resolution, and a comparison of Claude and ChatGPT (via API) behavior across 21 representative queries. The headlights category was selected because it spans the full range of query types found across aftermarket retail — the findings are intended to generalize to other parts categories, not to describe headlights specifically.
