Their rankings held steady. Their clicks went up 27×
The common fear about AI Overviews is that they take your traffic. This is a Singapore Shopify merchant whose category saw AI Overview coverage more than double across its category — and click-through rose ninefold in the same period, while the brand held its ranking positions throughout.
Challenge. Bread Garden already ranked. It was in Google's top five for 95% of the mooncake keywords we tracked — and still being passed over when Google answered the question itself.
Approach. Convert existing rankings into citations. Restructure the pages Google was already reading so the answer engine could extract, attribute and recommend them.
Result. Mooncake clicks rose from 34 in April to 918 in August. Click-through went from 0.48% to 4.41%. Average ranking position held, improving slightly from 10.9 to 9.7.
Takeaway. The traffic did not come from climbing the rankings — those were already strong and stayed strong. It came from being recommended rather than merely listed, while AI Overviews expanded across the category.
The Shopify store
Bread Garden is a Singapore bakery running a Shopify store for festive products — mooncakes for the Mid-Autumn Festival, goodies for Chinese New Year, pineapple tarts. Its commercial year is built around short, intense seasonal windows where most of the category's demand arrives in a six-to-eight week run.
Unlike a new store, it arrived with real advantages: an established domain, existing rankings, physical outlets and a customer base. Which is precisely what makes it the more useful case for most established Shopify merchants — the problem was not obscurity.
Ranked, but not recommended.
In May 2026 we audited 60 mooncake searches in Singapore. Bread Garden ranked in Google's top five for 57 of them and in the top three for 50. By any conventional SEO measure the work was done.
But the category was changing underneath those rankings. Google was increasingly answering mooncake questions itself rather than sending people to results: of those 60 keywords, 20 returned an AI Overview in May, and 44 did by July. More than twice as many questions were being resolved before a shopper reached the blue links.
That is the position most established Shopify merchants are now in, and it is more dangerous than being unranked, because nothing in a rank-tracking report looks wrong. You hold position three, the report is green, and the answer above you recommends someone else.
Converting a ranking into a citation.
The entity was already understood — Bread Garden is a known bakery with an established category. So the Brand Authority Directive work concentrated on the second half of the problem: making pages that an answer engine can lift from, attribute and recommend.
Separate the two scoreboards
Ranking and citation are different mechanisms, and conflating them wastes budget. Every tracked keyword was scored twice — where the site ranked, and whether the brand appeared in the AI answer. That split is what revealed the real gap: strong positions, missing citations.
Target where an answer already exists
Only keywords actually returning an AI Overview were worked. A keyword with no AI answer has no citation to win, and effort spent there produces nothing — the most common way seasonal GEO budget is wasted.
Make the pages extractable
Answer engines lift specific, checkable statements — flavours, certifications, prices, formats. Pages were restructured so the facts a shopper asks about sit in extractable form rather than buried in brand copy. The halal certification, the snow skin range, the durian varieties: each stated plainly enough to be quoted.
Work the seasonal window early
Mid-Autumn demand peaks in September. Citation is not instant, so the work has to land before the category heats up — the campaign was built to be in place by late June, ahead of the run.
Clicks up 27×, rankings held.
Mooncake queries only, from Bread Garden's own Google Search Console. April is the pre-campaign baseline. August is a partial month, to the 26th, and is already the strongest on record.
Where the traffic actually came from.
Read the bars against the last column of the table below them. Clicks multiplied by twenty-seven while average position held — improving by about a place, in a category getting more competitive by the week as the festival approached. Holding those positions through that is work in itself. But it is not where the clicks came from: if the gains had come from ranking movement, that column would show it.
| Month | Ranking queries | Clicks | Impressions | CTR | Avg position |
|---|---|---|---|---|---|
| Apr | 409 | 34 | 7,152 | 0.48% | 10.9 |
| May | 420 | 56 | 8,022 | 0.7% | 12.2 |
| Jun | 481 | 184 | 10,473 | 1.76% | 11.2 |
| Jul | 599 | 551 | 16,620 | 3.32% | 11 |
| Aug * | 781 | 918 | 20,826 | 4.41% | 9.7 |
* August to the 26th only. Mooncake-matching queries on the Bread Garden Search Console property, pulled 30 August 2026.
Named inside the answer, not buried in a source list.
Across a live audit of 60 keywords, 44 returned an AI Overview and 29 of those named Bread Garden. On 25 of them the citation is visible on screen without opening a source list — which is the difference between being a footnote and being a recommendation.
A few of the placements, with the organic position alongside, so the two scoreboards can be compared directly:
| Query | Vol / mo | Organic | In May |
|---|---|---|---|
| snowskin mooncake singapore | 200 | #1 | No AI Overview existed |
| best durian mooncake singapore | 80 | #2 | An answer showed, without Bread Garden |
| mooncake price | 80 | #1 | Cited |
| mooncake halal | 70 | #1 | Cited |
| black sesame mooncake | 40 | #3 | No AI Overview existed |
Live captures in Google Singapore, 10 August 2026, against a 4 May 2026 baseline. One annotated screenshot and one data file was kept for each of the 60 keywords audited.
best durian mooncake singapore is the clearest single result: in May an AI Overview existed for it and recommended other bakeries. By August it named Bread Garden, alongside Goodwood Park Hotel.
What this does not yet show.
The limitations
This is a seasonal category climbing into its peak. Some of the click growth is the Mid-Autumn window opening, not the work — which is exactly why the interesting number is CTR rather than clicks. Rising demand lifts impressions; it does not lift the rate at which people choose you.
Fifteen of the 44 AI Overviews still do not name Bread Garden. Average position is around nine, not three. There is real headroom on both scoreboards.
And a single seasonal cluster is not a whole business. This covers mooncakes; Chinese New Year and pineapple tarts are separate campaigns with their own windows.
What comes next
The uncited fifteen. Keywords already returning an AI answer where the brand is absent are the cheapest remaining wins — the answer exists, the ranking exists, only the citation is missing.
Gift and delivery queries. A distinct buying intent, adjacent to the cluster already won, and currently held by others.
The next festival window. Citation takes weeks to establish, so Chinese New Year work has to be in place well before demand arrives. Seasonal GEO is won in the quiet months.
AI Overviews did not take the traffic.
The fear is reasonable and widely held: Google answers the question, nobody clicks, the traffic disappears. In this category AI Overview coverage more than doubled over the period — from 20 of 60 keywords to 44 of 60. If the fear were the whole story, clicks should have fallen.
They rose twenty-seven fold, and the rate at which people clicked rose ninefold. Being named inside the answer turned out to be worth more than a blue link below it.
The distinction that matters is between being listed and being recommended. Bread Garden was already ranking — top five on 95% of the keywords audited — and held that through the season. Ranking was never the problem, and no amount of additional ranking work would have produced this. What changed is that the answer engine started naming it.
For an established Shopify merchant, that is the more common situation and the more urgent one: your rank-tracking report stays green while the answer above your result recommends someone else. Nothing looks wrong until you check the two scoreboards separately.
What this engagement involved.
Entity positioning (Brand Authority Directive) · ranking-versus-citation gap analysis · live AI Overview auditing across 60 keywords · Shopify content and collection restructuring for extractability · seasonal campaign sequencing · monthly citation and Search Console reporting.
Check your own AI visibility → See pricing → More case studies →
Published with Bread Garden's approval. Figures are from the client's own Google Search Console; AI Overview placements from live Singapore captures dated 4 May and 10 August 2026.