Gift Questions to AI: Is Your Brand Named This Q4?

How many people ask AI about gifts

The figure comes from a representative telephone survey of 1,002 people aged 16 and over, conducted in calendar weeks 41 to 46 of 2025 (Bitkom). Respondents were asked where they get ideas for their Christmas presents. AI lags well behind the familiar sources: 60 percent ask friends or family, 40 percent use suggestions from online shops and 26 percent use classic search engines (Bitkom). At 9 percent, the AI assistant is a small but distinct source (Bitkom). Unlike a brochure, it answers exactly the question that was asked.

More interesting than the share is the next step. 22 percent said they would find it "great" if an AI ordered the suggested gifts straight away (Bitkom). Outside Christmas, too, AI has reached product search: 16 percent of internet users aged 16 and over often use an AI chatbot when looking for a product, and among 16 to 29-year-olds the figure is 27 percent (Bitkom). These two values come from a second survey of 1,120 internet users in calendar weeks 40 and 41 of 2025 (Bitkom).

On top of this comes a broader shift in search itself. 50 percent of internet users at least sometimes use a chat with an AI instead of a classic web search (Bitkom). That matters for retail because online product research is part of everyday life: 80 percent of 16 to 74-year-olds searched online for information about goods and services in the three months before the 2025 survey (Destatis). And the gift itself is often bought online too: 73 percent of people aged 16 and over planned to order Christmas presents online in 2025 (Bitkom).

Many people will shop with the help of AI assistants in future. Retail has to prepare for the AI age now, otherwise there is a risk that its own offering will be found less often.

— Dr. Rohleder, managing director of Bitkom, press release of 24 November 2025 (translated from German)

Why a gift question works differently from a search query

A classic search returns a list of results that users click through themselves. A gift question to an AI assistant, by contrast, already contains the conditions of the purchase decision: for whom, for which occasion, up to what amount and deliverable by when. The answer condenses this into a few suggestions. The more precise the question, the narrower the selection. The German digital association Bitkom describes it from the perspective of the person asking:

The more specific the information on interests, hobbies or the available budget, the more accurate the suggestions become.

— Czachowski, AI expert at Bitkom, press release of 28 November 2025 (translated from German)

For retailers this means: measurement has to use questions that are as specific as the real ones. A check with the search term "tea gift" says little about whether the brand appears for "gift under 50 euros for a colleague who loves green tea". What matters then is whether your brand fits these conditions and says so explicitly on its pages.

Three characteristics of the answers add to this. First, links are often missing: in a Bitkom survey on searching with AI, only just over a third, 36 percent, find that the AI backs its answers with enough links (Bitkom). A mention can therefore have an effect without a click following. Second, answers contain errors: 42 percent of those who use AI for search say they have already received false or made-up information (Bitkom). Even a mention of your own brand can therefore include a wrong price, a discontinued product or a third-party seller.

Third, answers fluctuate. A study of five language models configured for repeatable output found accuracy variations of up to 15 percent across ten runs of the same tasks; none of the models delivered consistent results across all tasks (Atil et al., arXiv). A single test question in the browser is therefore not a measurement but a snapshot. The picture becomes reliable only when the same questions are repeated on a fixed schedule and read as a trend. How this works in principle is described in the article on GEO and measuring AI visibility.

The seasonal prompt set: collecting gift and purchase questions

The prompt set is the basis of measurement. In XICTRAQ AI visibility you define your brand, optionally aliases and competitors, plus typical customer questions, which can be purchase-intent, advisory or comparative. For the holiday season a separate profile next to the year-round one is worthwhile: the seasonal questions do not dilute the annual values, and in January the season can be evaluated on its own. The feature is included from the Business plan; how many profiles, questions per profile and queries per month are possible depends on the plan you choose (pricing and plans).

Good gift questions are not written at a desk; they come from what customers actually ask. There are three places to look: the search terms from last December's shop search, the questions sent to customer service and chat, and the landing pages through which gift buyers arrived last year. From these come question types that suit most retailers. The examples come from an imaginary tea mail-order shop:

• Budget question: "Gift idea under 50 euros for someone who loves tea". The budget narrows the selection considerably.
• Recipient question: "What should I give my father, who has just started drinking green tea?" It shows whether the brand is recommended to beginners.
• Occasion question: "Which advent calendar is worth it for tea drinkers?" It becomes relevant from late October, not only in December.
• Comparison question: "Your brand or Brand B as a gift?" It shows how the answer positions both providers.
• Source question: "Where can I order high-quality tea online as a gift?" This is the question that precedes the purchase.
• Deadline question: "Tea gift that still arrives before Christmas Eve". It matters in the second half of December.
• Voucher question: "Voucher or gift set for tea lovers?" Vouchers topped the popularity scale for Christmas presents in 2025 (HDE).

Whether your own brand appears in the question determines what the result says. Questions without a brand name show whether the brand is named unprompted; only they measure visibility. Questions with a brand name show how the answer describes the brand and belong in a small separate group. The unbranded case is the more important one: 28 percent of all internet users want to tell an AI what they need in future and let it find suitable offers (Bitkom). In that case the answer decides which brand is named, not the person searching.

As a working list from which the profile is created, the set can look like this:

profile: Gift questions 2026
brand: Your brand
aliases:
  - Your-Brand
  - YourBrand
competitors:
  - Brand B
  - Brand C
regions: [de, at]
schedule: weekly, daily in the sale weeks
questions:
  # without brand name: measure visibility
  - Gift idea under 50 euros for someone who loves tea
  - What should I give my father, who has just started drinking green tea?
  - Which advent calendar is worth it for tea drinkers?
  - Where can I order high-quality tea online as a gift?
  - Tea gift that still arrives before Christmas Eve
  - Voucher or gift set for tea lovers?
  # with brand name: show the positioning
  - Your brand or Brand B as a gift?

Two rules keep the measurement comparable. First, the wording of each question stays the same for the whole season. Rewording a question in November means measuring a different question from then on, and the trend breaks at that point. New questions are added as extra entries instead of replacing old ones. Second, brand and aliases are matched as whole words, regardless of upper or lower case. A spelling with a hyphen and one without are two aliases, and an alias does not count if it only appears as part of a longer word.

What is counted and what is not

Each run poses the prompt set to an AI model on a fixed schedule and logs the answers. Every answer is evaluated on four metrics that appear in the dashboard as a time series:

The regions Germany, Austria and Switzerland can be activated individually and are reported separately; the region is passed into the question as context. The prompt set applies per profile. If you sell in Switzerland and want to ask budget questions in francs there, it makes sense to create a separate profile for that. For each question, the individual results show whether the brand was named, how often brand and competitors appeared and, where the brand was named, which addresses the answer contained.

The trend from October to December

The holiday season is no side issue for retail: German retailers generate 18.5 percent of their annual revenue in the last two months of the year (HDE). For November and December 2025 the German Retail Federation expected revenue of 126.2 billion euros, of which 22.2 billion euros online (HDE). For Black Friday and Cyber Monday 2025 alone, the forecast was 5.8 billion euros (HDE). Customer questions shift during these weeks as well, so it pays to read the trend in phases.

In October the questions are about ideas and advent calendars, around Black Friday on 27 November 2026 about offers, and in December about delivery deadlines and vouchers. The sale days are gift days as well: of those who planned to shop on Black Friday or during Cyber Week in 2025, 33 percent said they would mainly use the discounts to buy presents for family and friends (Bitkom). Gift questions therefore belong in the sale week too, not just in December.

The dashboard helps compare the phases: a chosen period, such as the last 30 days, is set against the previous period of the same length. The time series in the profile shows the values per day, and the individual results can be filtered by region. This shows whether a drop affects all questions or only one group, such as the deadline questions in December.

You set the schedule per profile: daily, weekly or monthly. For the October baseline a weekly run is usually enough. In the sale weeks a daily run is more informative, because answers vary from run to run and a single weekly value may then come out by chance. Each question counts as a separate query per region against the plan's monthly allowance, so daily runs use it up correspondingly faster. Anyone who later compares the season with the previous year runs into the same problem as in web analytics: Black Friday moves around the calendar. The article on year-over-year comparison in Q4 shows how to compare sale week with sale week instead of date with date.

Start early: baseline and alert need lead time

The most important date is not in December but in October. Without values from before the promotions, there is no way to tell in November whether a mention rate is good or bad, because the comparison is missing. Then there is the alert: after each run, XICTRAQ compares the most recent individual queries of a profile with the ones immediately before them and reports when the citation rate has dropped clearly. Such a comparison needs queries from before a drop. A profile created only in Black Week knows only that week's values; a decline that set in earlier will no longer show up.

The alert goes out through the channels already set up for notifications, such as email, messenger or webhook (alerting and notifications). For it to have a target, the profile is linked to the website. Once the comparison no longer shows a clear drop, the alert resolves itself instead of staying open until January. That also happens when the rate stays low and the drop has moved out of the comparison window, so a resolved alert does not replace a look at the trend.

When the brand is missing: gift pages that can be cited

If the brand is missing from answers to gift questions, the lever is your own content. According to Google, there is no special optimisation for this: the usual SEO fundamentals apply to the AI features in Google Search, and there are no additional requirements (Google Search Central). Google also describes that these features can fan a question out into several related searches on subtopics (Google Search Central). A gift question thus breaks down into budget, recipient, delivery time and where to buy, and each of these sub-questions needs a page that answers it.

• Gift guide with budget levels: a guide with clear sections that explicitly state price range, occasion and recipient, instead of an image gallery without text.
• Consistent brand spelling: a shop that appears sometimes with and sometimes without a hyphen makes both answers and measurement harder. The variants also belong in the profile as aliases.
• Order cut-off and delivery notes: a clearly dated note on the last day to order for delivery before Christmas Eve answers the deadline question directly.
• Permanent addresses for seasonal pages: a gift guide that sits at the same address every year collects references over the years instead of starting from scratch. Why this applies to sale pages is shown in the article on the Black Friday page with a permanent URL.
• Access for AI crawlers: pages that AI crawlers are not allowed to fetch are usually not available for answers with source references. How to control and check access is described in the article on llms.txt and AI crawlers.

Such pages do not take effect overnight. XICTRAQ puts the questions to a model that answers from its own knowledge without searching the web live; a page published in November will therefore usually not appear in its answers within the same season. Whether AI crawlers fetch the page, on the other hand, can be measured separately, as the article on AI crawler traffic shows. There is no promise that a page will be named afterwards. If the individual results contain addresses that carry your brand name but do not belong to your domain, it is worth checking for imitators. The article on fake shops using your brand shows how to spot them before the sale week.

From mention to visit

A mention is not yet a visit, and a visit is not yet a purchase. Even with the AI summaries of classic search engines, 24 percent of internet users often rely on them without looking at the search results (Bitkom). Part of the effect of a mention therefore stays in the chat. The other part arrives as a visit, and web analytics measures that. How to recognise visits from AI assistants and follow them through to revenue is shown in the article on AI referral traffic and revenue.

In the holiday season it is worth placing both series side by side: the weekly mention rate from AI visibility and visits of AI origin, including orders, from cookieless analytics. If both rise in the same weeks, that is an indication, not proof, because promotions, weather and delivery deadlines act at the same time. If the mention rate falls while visits remain stable, it is more of an early warning than damage.

Set up gift questions as a prompt set now

The effort stays manageable if it starts in time. The following order puts the baseline before the promotions and prepares the January evaluation at the same time:

1. Collect questions: evaluate last December's shop search, customer service and landing pages and turn them into the seven question types.
2. Create the profile: brand with all spellings as aliases, the main competitors as terms, regions according to your sales markets.
3. Link and choose a schedule: connect the profile to the website so that the alert has a target, and start with a weekly run.
4. Read the baseline: by early November, check which questions name the brand, which do not and which pages appear as a source.
5. Close gaps: add gift guides, order cut-off dates and consistent spelling before the sale weeks.
6. Follow the sale weeks closely: measure daily between Black Friday and Christmas Eve and compare phase with phase.
7. Evaluate in January: set the season against the baseline and carry the prompt set over to 2027 with unchanged wording.

There is no promise of being named in AI answers. What can be measured is whether your brand appears for the questions customers actually ask in the holiday season, and whether that changes over the weeks. The live demo shows what the evaluation looks like without signing up.