How visible is Looff in AI answers? Two GEO analyses explained

Choosing a gift for a colleague no longer always starts in an online shop. Someone might ask an AI platform what suits a farewell, or which flowers fit through a letterbox. For a supplier, the question is not just whether a product can be found, but whether the website becomes a source in the answer.

Two supplied GEO analyses make this concrete for Looff and the domain looff.nl. One examines Cadeau Collega (colleague gifts), the other Bloemen en Planten (flowers and plants). They show a similar pattern: Looff is cited more often for action-oriented questions than for broad exploration. Yet results differ by topic and platform. This is an analysis of source reports, not a client case study or evidence of an optimisation result.

What was investigated?

The four original HTML reports carry the report date 14 April 2026, specify the Netherlands as the tracking location and use the platform labels ChatGPT, Google AI-mode and Perplexity. Google AI-mode here is not the same as Google AI Overviews. The source sets use both Loof and Looff in their labels, but both investigate looff.nl. We use Looff in this article.

Presence in these reports means the percentage of analysed AI answers in which a domain appears as a source at least once. Average source position uses the first, highest-placed citation of that domain in each answer. A lower position number means an earlier place in the source list. It is not a conventional Google ranking.

These figures measure source citations within selected prompts. They do not measure market share, revenue, search volume, website visits or brand mentions without a source link. The two studies use different questions and counts. We do not combine them into a single visibility score.

Answers are not the same as prompt rows

The platform statistics list 317 ChatGPT answers, 346 Google answers and 358 Perplexity answers for colleague gifts. For flowers and plants, the respective counts are 472, 602 and 483. These counts match the unique prompt texts per platform in the source exports. Multiple source rows within one answer are not multiple answers.

The intent tables show a distribution of 317 prompts for colleague gifts: 84 See, 75 Think, 94 Do and 64 Care. For flowers and plants they list 472: 116 See, 120 Think, 129 Do and 107 Care. However, the separate prompt files contain 376 and 602 data rows. The intent distribution is therefore not a universal denominator for every platform percentage. The exports do not establish why platform coverage differs. An answer not found in the data is not the same as an answer without a Looff citation.

The dates also require caution. Excel date values 46126 and 46127 correspond to 14 and 15 April 2026. Colleague gifts includes source rows dated 14 April for ChatGPT and Google and 15 April for Perplexity, while the domain export lists 15 April. Flowers and plants has 14 April in the inspected date fields. Whether these fields represent measurement time or export date is not documented. We therefore do not state a confirmed measurement period. This article's publication date is separate from the report date.

Comparing the two topics

The tables below retain presence to one decimal place and average source position to two decimal places, as in the KPI summaries. Small differences do not prove a lasting lead. These are selected questions, not all possible AI answers.

Google AI-mode cites Looff most often within both sets. For flowers and plants, that higher presence comes with a later average source position than on the other platforms. Perplexity gives Looff the earliest average source position in both studies, but not the highest presence. Being cited more often and appearing earlier in the source list are different things.

Colleague gifts: more visible for concrete choices

Ordering a farewell gift is a different question from finding out what is appropriate at work. The intent analysis distinguishes See for broad exploration, Think for considering and comparing, Do for taking action and Care for support and the relationship after purchase.

Do has the highest presence of all four intents on every platform. This makes concrete purchasing questions a logical starting point for further investigation. It does not prove that product pages are the cause, or that those citations generate orders.

At theme level, farewell stands out. For Do within Afscheid (farewell), the detailed KPI table lists 8.3% in ChatGPT, 23.1% in Google AI-mode and 7.7% in Perplexity. The label lists 12 prompts, but that is not an equal platform denominator: the Google and Perplexity percentages reflect different coverage. Treat this small segment as a research signal, not an overall score for farewell gifts.

The overview table includes, for example, Looff's page about gifts for colleagues who are ill. It also contains pages about coffee and tea gift packs and delivery to multiple addresses. Other sources include Appical's article about welcome packs and YourSurprise on farewell gifts. Competition for a source position is not limited to similar online shops. Explanations and practical decision support also appear.

Flowers and plants: product type matters

Here too, Do is higher than See, Think and Care on every platform. However, exploration differs from colleague gifts: Google AI-mode lists 7.3% for See, while ChatGPT shows 0.0%. One platform cannot speak for the whole AI landscape.

* Zero is a report display here. The HTML summaries show 0.0% for these combinations. In the intent worksheet, the corresponding domain and KPI cells are blank; the detailed theme tables contain explicit zeros. We read this as no recorded Looff source citation within the reported selection, not proof that every possible answer was measured. A displayed average position of 0.00 when absent is not a real position and has not been reproduced.

The themes range from small trees and letterbox flowers to dried flowers, felt flowers and occasion bouquets. For Do within brievenbusbloemen (letterbox flowers), Looff is at 6.2% in ChatGPT, 10.5% in Google AI-mode and 0.0% in Perplexity. For See within viltbloemen (felt flowers), all three detailed KPI rows list 0.0%. That is not the same as saying Looff is invisible for flowers everywhere.

Verifiable Looff sources include Season Flowers letterbox flowers and the wedding tree. The source exports also contain domains such as BloomPost and Viltbloemist. These are examples of cited parties, not a market-share ranking. These historical source links may since have changed.

From observations to practical next steps

Stronger Do presence might fit the usefulness of concrete offerings, but the reports do not test that explanation. They also do not show why a platform selects a source. The recommendations below are therefore editorial interpretations to test, not proven causes or promises of growth.

1. Start with cited pages. Check whether information about gift contents, personalisation, delivery and ordering is clear and current. Retain useful explanations when improving pages. A source citation is a reason to assess a page, not a quality certificate.
2. Support exploration. Address questions such as what is appropriate for a farewell or illness, and when letterbox flowers, a bouquet or a small tree are suitable. Explain the basis for advice and the situation it applies to.
3. Make comparison easier. Present differences in size, care, personalisation and delivery clearly alongside each other. Connect decision support to relevant categories and products, rather than only creating a commercial product list.
4. Do not forget Care. Investigate which questions about care, delivery and help after purchase remain insufficiently answered. Low recorded presence is a reason to examine that coverage, not proof that existing customer service falls short.
5. Make the next comparison reproducible. Define prompts, platforms, coverage, date meanings and the treatment of missing answers in advance. Only then compare a new measurement using the same design. Assess traffic and enquiries separately; they cannot be inferred from these citation percentages.

View the underlying reports

The report documents below are in Dutch and carry the report date 14 April 2026. Open a report to explore the original tables and interactive elements. Charts depend on external scripts. If those fail, the main conclusions and comparison tables above remain readable. In the public copy of the colleague gifts overview, a test page and a client-specific product reference have been masked. Other figures and the original attachments have not been changed. Raw Excel files are not published.

What does this mean for your visibility?

These two analyses show why a general AI score tells you too little. Break the question down by topic, intent and platform, check what was actually measured and examine the cited pages. That creates an evidence-based priority list rather than an unproven promise.

Want to explore conventional search visibility and visibility in AI answers together? Read more about SEO & GEO or get in touch to discuss your question.