Skip to content

Reducing single-engine dependence for a fragrance retailer

Almost all of TryScent's AI visibility rested on one engine, and nobody had measured that. When ChatGPT's coverage of the brand later contracted, the store did not vanish from AI answers, because by then three other engines were citing it.

Company
DTC e-commerce, fragrance
Category
Fragrance retail
Duration
7 months
Services
Technical SEO, AEO, GEO
Results after seven months
68%
Of AI citations now come from outside ChatGPT, up from 2%
4 of 7
AI engines citing TryScent, up from 3
+90%
Referring domains, 147 to 280
+20%
Domain Rating, 20 to 24

Measured in Ahrefs against a fixed January 2026 baseline. Total citation volume fell over the same period as ChatGPT's coverage of the brand contracted; the results section sets that out in full.

01

Problem

TryScent sells fragrance direct to consumers across men's, women's and unisex collections. Like most stores of its size it was running on a theme that had never been audited, and the technical debt had compounded quietly for as long as the store had existed.

A page-level audit found over three thousand technical errors across crawlability and site structure. Schema markup was missing across the store, which cost rich result eligibility and, more importantly, left AI crawlers inferring product attributes rather than reading them. URL slugs and product pages did not match in places, which interferes with how products get indexed at all. Domain authority was weak.

The finding that reframed the engagement came from splitting AI visibility by engine rather than reading it as one number. Ninety-eight percent of TryScent's AI citations came from ChatGPT alone. On paper the brand had AI visibility. In practice it had a single point of failure that nobody had looked at closely enough to notice.

02

Baseline

The January 2026 snapshot combined an Ahrefs pull with a page-level technical audit. Read together, the two halves explained each other: a store engines could not parse cleanly, and a citation base resting entirely on one of them.

3,000+
technical errors across crawlability and structure
98%
of AI citations came from a single engine
0
product pages carrying schema markup
20
Domain Rating on a weak authority profile

Engine coverage tracked across ChatGPT, Perplexity, Gemini, Google AI Overviews, Google AI Mode, Copilot and Grok. Error count and schema coverage are audit findings rather than Ahrefs metrics.

03

Strategy

Two problems, one of them urgent and one of them structural. The store needed to be readable, and the brand needed to stop depending on a single engine for its entire presence in AI answers. Three moves.

A

Turn three thousand errors into a sequence

A page-level audit converted into a prioritised cleanup roadmap, then worked through in that order. Three thousand items attacked at random goes nowhere. Ten in the right order make the next ninety possible.

B

Restore the structured data layer

Schema across product pages so the store was eligible for rich results and legible to AI crawlers reading notes, categories and price as data rather than inferring them from page copy.

C

Spread the citation base deliberately

Content and formatting aimed specifically at the engines not already citing the brand. Visibility that rests on one engine is not visibility, it is exposure that happens to be pointing the right way.

And what we deliberately did not do

We did not build a strategy around defending the ChatGPT position. Citation volume on any single engine moves for reasons entirely outside a brand's control, and organising the work around holding one number would have left the store more exposed rather than less.

04

Execution

SearchAxe ran the remediation and the structured data rebuild. The technical work went first, because schema and content land on nothing if the crawler cannot get through the store reliably.

  • Ran a page-level audit converting three thousand technical errors into a prioritised, sequenced cleanup roadmap.

  • Worked through remediation across crawlability, site structure and product page architecture.

  • Fixed URL slug and product page mismatches interfering with how products were being indexed.

  • Implemented the missing structured data across the store to restore rich result eligibility.

  • Made product attributes machine-readable so AI crawlers could read them rather than infer them from copy.

  • Produced blog content built for AI Overviews, AI Mode and LLM extraction.

  • Ran link acquisition to strengthen a thin authority profile.

05

Results

The structural change is the one worth reading. Seven months ago the brand's presence in AI answers was effectively one engine's decision. Today four engines cite it, and two thirds of those citations come from engines that contributed almost nothing at the start.

Metric
Jan 2026
July 2026
Share of AI citations outside ChatGPT
2%
68%
AI engines citing TryScent
3 of 7
4 of 7
Perplexity citations
1
33
Copilot citations
2
25
Grok citations
0
5
Referring domains
147
280
Domain Rating
20
24
Keywords on page one
1
4
URLs stranded beyond position 50
18
0

Source: Ahrefs. July 2026 used as the last complete month; AI citation counts captured August 2026. Stated in full: total citation volume fell from 140 to 92 across the period, because ChatGPT citations for the brand contracted from 137 to 29 while the other engines grew. Organic traffic value also declined over the same window, and the store's organic footprint remains small in absolute terms. The engagement was scoped to technical health, structured data and citation breadth rather than traffic volume.

06 · The part that transfers

Lessons for e-commerce brands

Concentration is invisible until it breaks

Ninety-eight percent of TryScent's AI citations came from one engine, and no one had measured it because AI visibility was being read as a single number. One number hides the risk sitting inside it.

Three thousand errors is not a to-do list

An audit that produces three thousand items produces paralysis, not progress. Sequencing the list into something that can actually be started on Monday is the deliverable. The list is just the input.

For a store, schema is product data

Structured data is how an engine reads notes, sizes, categories and price without guessing. Missing schema does not only cost rich results. It costs accuracy in the AI answers that describe your products to buyers.

Engine coverage moves on its own

ChatGPT citations fell while Perplexity and Copilot rose over the same months. Track engines separately or you will misattribute both your wins and your losses to whatever you happened to ship that quarter.

Ready to turn search into a growth channel?

Let's build a stronger organic growth engine.

hi.searchaxe@gmail.com All case studies