Why a gift question is not a search query

A keyword search returns a page of options and leaves the filtering to the shopper. A gift question carries the filters inside it: a recipient, a constraint, a budget. Google describes shopping in AI Mode in exactly those terms: describe what you are looking for “just as you’d say it to a friend” (Google, 13 November 2025). The assistant narrows the field and names a few products.

Search query and gift question

One asks for a list, the other asks for a pick

A keyword query hands the shopper a results page to filter. A gift question carries the filters inside it, so the assistant does the narrowing and names a few products.

Hypothetical example

Keyword search

gifts for dad

One broad topicNo budgetNo constraint

A results pageThe shopper opens several stores and does the filtering.

Gift question to an assistant

gift for a dad who has everything, under 100

RecipientDadConstraintHas everythingBudgetUnder 100

A short list of named productsThe assistant does the filtering, using the product data and text it can read.

The constraints in the question become the filters. A product whose data and page text do not state them is hard to match.

For a store that runs paid ads, this changes what can be bought. A position on a results page can be bid for. The products named in an answer cannot, at least in ChatGPT: OpenAI states that “product results are selected independently by ChatGPT and are not ads, nor influenced by any OpenAI partnerships”, and that ads are separate from product results. The lever is the information the assistant can read about your products, not budget.

The second half of the problem sits on your own website. A shopper who clicks through from an answer has already been told which product fits and why. If the page they land on shows a different price, no delivery date, or a generic welcome offer, the recommendation is wasted. The purchase is the conversion event here, and the measure is revenue per visitor from that traffic, not the number of mentions.

How products come to be named in ChatGPT

ChatGPT, per OpenAI

From question to named product

What OpenAI’s help article says happens between a shopping question and the products shown. Read on 5 October 2026.

1

Question and context

The query, plus context such as Memory or Custom instructions.

2

Relevance to intent

• Structured metadata from first-party and third-party providers, such as price and description
• Model responses generated before new search results
• Safety standards and product policies

3

Products shown

Selected independently. Not ads. Review summaries come from reviews on public websites.

4

Merchant list

Ranked on factors like availability, price, quality, and whether the merchant is the maker or primary seller.

Every input in step 2 is something a store publishes or earns. None of it is a placement that can be bought.

Two details matter for a gift season. First, data lags: OpenAI notes that when merchants update pricing or shipping terms “there may be some delay before it is reflected”, so a promotion launched the same morning may not be what the shopper is told. Second, the sale happens on your side. The merchant page states that “purchases are completed on merchant-owned websites or apps”. An assistant can only recommend what it can read, and it can only read what a store has published as data or as plain text.

How products come to be named in Gemini and Google AI Mode

Google’s position is that there is nothing separate to optimize. Its guide to generative AI features in Search (updated 10 July 2026) says that “optimizing for generative AI search is optimizing for the search experience, and thus still SEO”. To be a supporting link in AI Overviews or AI Mode, a page must be “indexed and eligible to be shown in Google Search with a snippet” (updated 10 December 2025).

What is different is how the question is handled. Both features may use query fan-outOne question is split into several related searches, and the results are combined into one answer., which Google describes as “issuing multiple related searches across subtopics and data sources”. A gift question therefore touches more of your catalog and content than a single keyword would.

Gemini and Google AI Mode, per Google

One question, several searches

Google says AI Overviews and AI Mode may use query fan-out: related searches across subtopics and data sources. The sub-searches below are a hypothetical example. The data sources are the ones Google documents.

The questiongift for a dad who has everything, under 100

Fan-out sub-searchesHypothetical examplegifts for someone who has everythingexperience gifts under 100best rated gifts for dads

Data sources Google documents

Indexed web pagesIndexed and eligible to show with a snippet.

Shopping GraphMore than 50 billion product listings, 2 billion updated every hour.

Merchant Center and Business ProfileKept up to date by the store.

Each sub-search is a chance for a different page or listing of yours to match. Shopping Graph figures: Google, 13 November 2025.

Product data comes from the same place as Shopping results. Google’s guide says that “using products like Merchant Center (such as Merchant Center feeds) and Google Business Profiles can help your products and services to be visible in both AI responses and other Google Search results”.

The Shopping GraphThe product database Google uses for shopping results, built from merchant feeds and the web. sits behind that data. Google put it at “more than 50 billion product listings, 2 billion of which are updated every hour” when it announced shopping features in the Gemini app on 13 November 2025.

On structured data, the guide is careful: it “isn’t required for generative AI search, and there’s no special schema.org markup you need to add”, but it is worth keeping as part of normal SEO, and it has to match the visible text. On third-party mentions it is blunt: “Seeking inauthentic ‘mentions’ across the web isn’t as helpful as it might seem.” For AI Overviews specifically, see how to optimize for Google’s AI Overviews.

The four inputs, side by side

Put the two sets of documentation next to each other and the same four inputs appear in both. The table states only what each company has published.

Sources: OpenAI Help Center, merchant page, feed specification and crawler documentation, read 5 October 2026. Google Search Central, updated 10 December 2025 and 10 July 2026.

An assistant can only recommend what it can read, and it can only read what a store has published as data or as plain text.

Neither company promises an outcome. Google’s guide says that meeting every requirement “doesn’t mean that Google will crawl, index, or serve its content”, and OpenAI describes selection as the model’s own judgment of relevance. For brands that have never ranked for these questions, how to show up in an AI answer without ranking covers the retrieval side in more depth.

What a store can do in the weeks before the gift season

Four weeks is enough for the work that depends on you. It is not enough to guarantee a result, because crawling, indexing and selection are not yours to schedule. Do the steps in this order.

Four weeks before the season

Fix the data first, then the pages

The order matters. Prices, stock and access problems block everything after them, and review and content changes need time to be crawled.

Week 1Data and access

• Price and stock match across feed, structured data, page and checkout
• Identifiers, titles and descriptions complete
• Crawlers allowed

Week 2Answerable pages

• Who the product is for, in text
• Price band, delivery cut-off, returns
• Gift collections by recipient and budget

Week 3Proof

• Recent reviews on product pages
• Reviews on public sites
• No manufactured mentions

Week 4 onwardsCheck and arrival

• Fixed prompt set, re-run weekly
• AI-referred sessions segmented
• Revenue per visitor tracked

Outside the gift season the same sequence works as a quarterly routine.

1. Reconcile price and availability. Google’s product data specification requires the price to “match with the price from your landing page, structured data, and at checkout”. The same discipline protects you in ChatGPT, where stale data is what the shopper gets quoted.
2. Complete the attributes. A GTINGlobal Trade Item Number: the standard barcode number that identifies a product across retailers. where one exists, plus titles, descriptions, images and brand. Gift questions filter on attributes, so a missing one is a missed match.
3. Check access. Confirm robots.txt does not block OAI-SearchBot or Googlebot, that product and category pages are indexed, and that your catalog reaches Merchant Center and, where available to you, ChatGPT.
4. Write the answer into the page. State in text who the product suits, the price band, the last order date for delivery, gift wrapping and returns. Build gift collections by recipient and budget. Google asks for important content in textual form, and an image of a delivery banner is not text.
5. Collect real reviews. Ask recent buyers, on your product pages and on the public review sites your category uses. Do not buy mentions.
6. Set up the check and the arrival. Fix a prompt set, segment AI-referred sessions in analytics, and decide what those shoppers should see first. The next two sections cover both.

How to check whether your store is being named

Use a fixed prompt set, run it more than once, and count shares. A single answer tells you almost nothing, because answers vary. ChatGPT takes context such as Memory and Custom instructions into account, Google’s fan-out searches can differ between runs, and prices and stock move underneath both.

Checking visibility

One answer is an anecdote, a prompt set is a measure

Five fixed prompts, each run three times in one assistant. The store is named in some runs and not in others, so the useful number is the share of runs, tracked week over week.

Hypothetical example

PromptRun 1Run 2Run 3

P1gift for a dad who has everything, under 100NamedNamedNot named2 of 3

P2last minute gift for a colleagueNot namedNot namedNot named0 of 3

P3gift for a new homeowner under 50NamedNot namedNamed2 of 3

P4anniversary gift that is not jewelryNamedNamedNamed3 of 3

P5gift for a teenager who likes cookingNot namedNamedNot named1 of 3

Named in 8 of 15 runs this week. Compare with the same 15 runs next week.

Keep the wording of each prompt fixed. Changing the prompt changes the test.

• Write 15 to 20 prompts in shopper language. Combine recipient, budget and occasion. Take the wording from site search, customer service and reviews, not from keyword tools.
• Keep the wording fixed. If the prompt changes, the comparison with last week is gone.
• Run each prompt several times per assistant from a clean session, so your own history does not shape the answer.
• Record four things: whether you were named, which product, which competitors, and which sources were linked.
• Repeat weekly and compare the share of runs, by prompt group.

Then look at the traffic. Google reports clicks from AI features inside the Web search type in Search Console, and its guide points to a generative AI performance report there. For ChatGPT and other assistants, segment sessions by referrer in your analytics. Prompt checks show whether you are visible. Only the sessions show whether it is worth anything.

What happens on the website when an AI-referred shopper arrives

Revenue per visitor

The higher conversion rate is not always the better traffic

Revenue per visitor is conversion rate multiplied by average order value. Two traffic sources can swap places once order value is included.

Hypothetical example

Source A

3.0%Conversion rate

×

$60Average order

$1.80revenue per visitor

Source B

2.0%Conversion rate

×

$110Average order

$2.20revenue per visitor

Source A converts half as often again, yet each visitor from Source B is worth 22% more. Judge AI-referred traffic on the second number.

Three things on the page decide whether the visit pays:

• Land on the product that was named, with the price, stock and delivery date the answer implied visible without scrolling.
• Answer the gift constraints again: last order date, gift wrapping, returns after the holiday.
• Do not treat the visitor as cold traffic. A first-visit discount popup interrupts someone who arrived ready to check out.

This is the part Pathmonk works on. It reads each visitor’s behaviour in the session, without third-party cookies, and adapts the conversion layer of the page to match: the proof, the offer and the next step. A shopper who arrives from an AI answer and goes straight to delivery details is shown something different from one who is still comparing. Each experience is measured against a control group, so the effect on purchases is measured, not assumed.

Key takeaways

• A gift question carries its own filters. The assistant narrows the field and names a few products, using data and text it can read.
• ChatGPT product results are not ads, according to OpenAI. Visibility comes from product data, reviews and crawlable pages.
• Google treats AI Mode and AI Overviews as Search. No AI text files or special markup are needed, and Merchant Center data helps.
• Price and availability must match across feed, structured data, page and checkout. Stale data is what the shopper gets quoted.
• Write who the product is for, the budget band and the delivery cut-off in plain text on the page.
• Measure visibility with a fixed prompt set, several runs each, repeated weekly. Single answers vary too much to trust.
• Judge AI-referred traffic on revenue per visitor, and make the landing page confirm what the answer promised.
• No tactic guarantees a mention. Both OpenAI and Google say so in their own terms.

FAQs on gift questions and AI product recommendations