How many leading gift searches showed an AI Overview?#
98 of the 100 leading gift-search phrases showed an AI Overview in Ahrefs’ Google US snapshots dated 20 to 26 October 2026. Lily AI ranked the phrases by estimated December 2025 US search volume. This is a dated search-result observation, not a claim about every shopper’s experience or the share of purchases influenced by AI. [A1]
- For 25 years, the keyword decidedIt decided which products a shopper saw: in the ad auction, in the site search box and in the catalog written to match both.
- This holiday, it stops decidingFor the requests that fill gift season, Google now answers with AI, built from what pages and listings state.
- Keep the keyword for known itemsBrand, model and reorder searches still work. Give the listing a budget line and an owner for everything else.
- 98 / 100
- leading gift phrases showed an AI Overview.Last December’s top 100 US gift searches, in Ahrefs’ records of Google, 20 to 26 October 2026.
- 98.1%
- of December gift-search volume named no product.Share of December 2025 searches across 598 gift phrases (Ahrefs US estimates).
- 43%
- of benchmarked sites had use-case search issues.Share of 170+ sites in Baymard’s 2026 benchmark, against 12% for exact product searches.
Sources: Lily AI analysis of Ahrefs US estimates and of Ahrefs’ records of Google’s US results, October 2026; Baymard Institute, updated 29 April 2026. An AI Overview is an answer on the results page, not a sale; the measurement section sets out what to measure. [A1][A2]
How did AI Overview coverage change from 2025 to 2026?#
Among the same 30 leading phrases with historical records, recorded AI Overview presence rose from 0 of 30 in October 2025 to 10 of 30 around December 2025 and 30 of 30 in October 2026. In the larger matched group, presence rose from 34 of 94 to 92 of 94. [A1]
Wide table: swipe horizontally or focus the table and use the arrow keys to see every column.
| Group and snapshot | Available records | With AI Overview | Share of phrases | Share of Dec 2025 volume |
|---|---|---|---|---|
| Top 30 / October 2025 | 30 | 0 | 0.0% | 0% |
| Top 30 / December 2025 | 30 | 10 | 33.3% | 26.7% |
| Top 30 / October 2026 | 30 | 30 | 100.0% | 100% |
| Top 100 / December 2025 | 94 | 34 | 36.2% | 29.3% |
| Same 94 / October 2026 | 94 | 92 | 97.9% | 99.1% |
| Top 100 / October 2026 | 100 | 98 | 98.0% | 99.1% |
Ahrefs was already recording AI Overviews in October 2025: its 19 October 2025 snapshot of “what is a white elephant gift” shows one. [A1]
Six phrases in the top 100 named a year and lacked a historical December snapshot. The matched 94-phrase comparison avoids treating those missing records as if they showed no AI Overview. The historical top 30 are ranked within that 94-phrase group, excluding three year-naming phrases that would otherwise enter the top 30.
The top 100 phrases account for 8,898,069 of the full dataset’s 12,939,156 December 2025 searches, or 68.8%. In October 2026, the 98 phrases with an AI Overview represent 99.1% of the top 100’s estimated December search volume. 98% is a share of phrases; 99.1% is a volume-weighted share.
Of the current top 100, 98 show an AI Overview. The two that don’t: “christmas gift baskets”, which names a product, and “gifts for 8 year old boys”. The latter had an AI Overview in an earlier October check and lacked one in a later check, illustrating why the date and observation method matter. Google’s unpaid product listings also appeared on 63 of the top 100 result pages in the same October 2026 records. [A1]
Shoppers asked for what they asked for the year before. What changed is the system answering them.
What do shoppers specify when they have not chosen a product?#
They often specify a person or occasion. In the 598-phrase dataset, 59.7% of December 2025 search volume named a person and 58.5% named an occasion. The tags overlap: one request can name a person, an occasion and a budget at the same time. [A1]
Wide table: swipe horizontally or focus the table and use the arrow keys to see every column.
| The search names | Phrases | Estimated searches | Share of volume |
|---|---|---|---|
| A person | 407 | 7,729,004 | 59.7% |
| An occasion | 419 | 7,574,850 | 58.5% |
| A judgment word | 154 | 2,273,848 | 17.6% |
| A budget | 41 | 567,903 | 4.4% |
| A passion or role | 44 | 507,823 | 3.9% |
| A product | 20 | 245,901 | 1.9% |
| A situation | 19 | 206,572 | 1.6% |
578 phrases, accounting for 12,693,255 searches, named no product. The remaining 20 phrases accounted for 245,901 searches and were tagged as naming a product, including gift baskets and tech gifts. Only that product/no-product split is mutually exclusive.
The head of the list makes the distinction concrete. “White elephant gift ideas” drew an estimated 964,875 searches in December 2025, and “gifts for runners” 24,618. Each leaves much of the product decision to the system answering the request. [A1]
A product title that identifies the brand and item type can be accurate while leaving those requests unanswered. The missing information may be activity, age, fit, compatibility, gift suitability or a reason to prefer the item over a cheaper alternative.
Are gift requests becoming more specific?#
In a tracked cohort of 232 “gifts for” phrases, November and December search volume for phrases of five or more words grew 52.1% from 2019 to 2025, compared with 11.9% for three-word phrases. The rise paused in 2025, so the data supports a multi-year pattern rather than a sudden 2026 behavioral shift. [A1]
| Measure | Change in estimated searches |
|---|---|
| Three words | 11.9% |
| Four words | 14.5% |
| Five or more words | 52.1% |
One real search, growing
“gifts for women who have everything”
12,824 November and December searches in 2019. 29,576 in 2025 (Ahrefs US estimates). [A1]
The growth came mostly from exact ages. In AI Mode the requests run longer still: on 28 October Alphabet said AI Mode queries average three times the length of traditional searches. Write the facts that stay true: the age range, the activity, the material, the fit and what it works with.
Source: Lily AI analysis of Ahrefs US estimates, November plus December searches, 2019 to 2025, for the 232 “gifts for” phrases already tracked in 2019, pulled 20 October 2026; the rise came mostly between 2022 and 2024 and paused in 2025. Alphabet Q3 2026 earnings remarks, 28 October 2026. [A1][A6]
Why do use-case requests expose gaps in onsite search?#
A use-case search requires the retailer to connect a need with products that meet it. Baymard’s 2026 benchmark found issues with use-case searches at 43% of sites, compared with 12% for exact product or model searches. [A2]
| Measure | Sites with issues |
|---|---|
| Exact product or model | 12% |
| Product type | 20% |
| Symptom | 37% |
| Feature | 39% |
| Use case | 43% |
| Compatibility | 44% |
| Abbreviation or symbol | 54% |
| Non-product | 66% |
The benchmark covered more than 170 sites and apps and more than 10,000 performance ratings. Separately, Baymard found that 56% of sites failed to support shoppers’ searches adequately overall.
Google’s own example: the machine writes its own keyword
How Google describes AI Mode handling one shopping request.
A cute travel bag. Then: for a trip to Portland, Oregon, in May.
What AI Mode does
- Runs several searches at once
- Works out what makes a bag good for rainy weather and long journeys
- Suggests waterproof options with easy-to-reach pockets
“Waterproof”: the shopper never typed it. The machine wrote it. Only listings that say it can match.
A keyword match checks one word at a time. A request arrives as a set of conditions, and a machine can only check the ones a listing states.
Source: Google, AI Mode shopping announcement, 20 May 2025, which calls this step a query fan-out. Google’s example, not a Lily AI test. [A6]
Which product facts make a descriptive gift request answerable?#
Lily AI recommends a five-part brief: who the product suits, what it is for, the occasion, material and fit, and the evidence for its value.
Wide table: swipe horizontally or focus the table and use the arrow keys to see every column.
| Fact | What last December’s gift searches show | Question it can answer |
|---|---|---|
| Who it suits | Named in 59.7% of last December’s gift searches. A field can record a gender and an age band, not a mother or a teacher. | Is it suitable for this person? |
| What it is for | A passion, a use or a situation: runners, coffee lovers, always cold. Named outright in 5.5% of the searches, and worked out from context by AI Mode, as Google’s travel-bag example shows. | Will it support what they do? |
| Occasion | Named in 58.5%: Christmas, a white elephant exchange, a stocking, a secret santa. Google and Meta have no dedicated field for it. | Does it make sense for this gift occasion? |
| Material and fit | The one fact the specifications were built for. What fit means in use, such as waterproof or runs small, still lives in free text. | Will it fit, work or feel right? |
| Why it is worth more | 17.6% of the searches asked for a judgment, with words such as best, unique or funny. No specification has a field for it. | What justifies choosing this over an alternative? |
Shares of December 2025 gift searches, Ahrefs US estimates with Lily AI tagging; one search can name several facts. The list of five is Lily AI’s synthesis of the searches and the platforms’ specifications. [A1]
Use verified evidence: a manufacturer’s specification, an approved product record or the retailer’s actual policy. An enrichment should not invent a certification, review score, compatibility claim, delivery promise or customer outcome.
A useful internal check is: “Which statement in this listing answers the shopper’s condition, and where did that statement come from?” If the team cannot point to the evidence, the listing is not ready to support that claim.
Do standard product feeds contain all five facts?#
Three of the five facts have no dedicated field at Google or Meta. In the product specifications as published on 29 October 2026, broad audience fields such as gender and age group existed, while occasion and use-case facts depended on descriptive text or category-specific attributes. [A3]
Wide table: swipe horizontally or focus the table and use the arrow keys to see every column.
| Fact | Google Merchant Center | Meta catalog | Shopify taxonomy |
|---|---|---|---|
| Audience | Gender and age group | Gender and age group | Target gender and age group |
| Activity or use | No dedicated field | No dedicated field | Activity in some categories |
| Occasion | No dedicated field | No dedicated field | Occasion, occasion style or theme, in some categories |
| Material and fit | Material, pattern, size, color and related fields | Material, pattern, color, size | Varies by category |
| Evidence of value | Price fields only | Price fields only | Price and compare-at price only |
Google has started to patch the gap
Its question_and_answer attribute, built mainly for conversational experiences such as AI Mode, takes up to 30 questions and answers per product. Its guidance: “Avoid listing keywords or search terms.” [A3]
Source: Google Merchant Center product data specification and question_and_answer help page, Meta catalog field reference and Shopify’s open product taxonomy (release 2026-10), checked 29 October 2026. Custom and internal labels are not counted. [A3][A7][A8]
What happens when competing listings have the same facts?#
In one controlled skincare study, three AI models selected the known brand in all 670 valid baseline trials when specifications and prices were identical. A separate gradient test within the study found that relatively small rating or price advantages could change selections. [A4]
Wide table: swipe horizontally or focus the table and use the arrow keys to see every column.
| Measure | Reported result | Scope |
|---|---|---|
| Known-brand selection with matched facts | 670 of 670 valid trials | Baseline experiment 1a, three models |
| Price advantage for half the selections | 7.3% lower price | Interpolated threshold, experiment 1c |
| Rating advantage for half the selections | 0.075 stars higher | Alternative interpolated threshold, experiment 1c |
The models were GPT-4o-mini, Claude Sonnet and Gemini 3 Flash. The thresholds were estimated by linear interpolation from experiment 1c’s 9,220 trials; the 670-trial baseline was experiment 1a. The work is an independent preprint by Xi Chu and Yupeng Hou, revised 10 October 2026, not peer reviewed. The models and controlled product setup limit generalization.
Retail interpretation: state real differentiators and test their effect. The study is not a pricing rule, a guaranteed AI recommendation formula or a Lily AI performance result.
Does this mean retailers should stop investing in keyword search?#
No. Known-item searches, brand and model queries, and working paid-search campaigns still have value. The change is an additional need for product information that can answer a descriptive brief.
Alphabet’s earnings release for the third quarter of 2026 reports Google Search and other revenue of $56,567 million in Q3 2025 and $63,350 million in Q3 2026, up 12.0%. “Search and other” includes Google’s other properties, so it is broader than search ads. [A5]
| Measure | US$ billions |
|---|---|
| Q3 2025 | 56.567 |
| Q3 2026 | 63.350 |
Google says the keyword is giving way
- Google says early AI Mode testers ask questions two to three times the length of traditional searches.
- Google starts rolling out agentic checkout with eligible US merchants: it buys a tracked item when the price fits the shopper’s budget, after the shopper confirms.
- Sundar Pichai tells the National Retail Federation that shopping journeys are moving “from keywords to natural conversations”.
- By July, AI Mode had passed 1 billion monthly active users, against more than 100 million in the US and India a year earlier (Alphabet, company-reported).
- AI performance insights for AI Mode and AI Overviews reach US Merchant Center businesses. They cover organic results only.
- On 15 October, Google begins rolling out Universal Cart in the US, with price-drop notifications.
Sources: Google I/O post, 20 May 2025; Google, 13 November 2025; Sundar Pichai’s NRF remarks, 11 January 2026; Alphabet Q2 2026 earnings remarks, 22 July 2026; Google, 16 September and 15 October 2026. User counts are company-reported. [A6]
Keep the search budget. Then give the listing what the keyword has had for 25 years: a budget line, an owner and a weekly report.
What should retail teams change first?#
Start with priority products and recurring category questions. Record what the listing can answer, which fact is missing, who owns the fix and when it publishes.
Wide table: swipe horizontally or focus the table and use the arrow keys to see every column.
| Owner | Before your freeze | During the freeze |
|---|---|---|
| CMO or commerce lead | Check what Google shows for the ten gift searches from last December your products should answer, and whether they appear. | Review Merchant Center’s AI performance report in the same weekly meeting as the search terms report. |
| Category team | Write the five facts in plain words for your top 100 holiday products. | Keep a list of every site search that returns nothing useful. It is the brief for January. |
| Feed and content team | Put the facts where machines read them: product highlights, product details, question_and_answer, the description and the page. | Make additions through feed sources that need no site release, and keep the page and the feed saying the same thing. |
| Performance team | List which Search campaigns run AI Max, and read the landing pages it can use to write ad text. | Each week, tag the search terms matched beyond your keyword list by the facts they ask for, and send the gaps to the catalog team. |
| Analytics team | Agree one measure before holiday traffic arrives: the share of last season’s top requests your listings can answer. | Measure listing changes against a control, and report the share monthly next to search spend. |
In January
Open a budget line for product content beside the search budget, with an owner and a weekly report. Add the five facts to item setup, so every new product arrives with them.
How should a retailer measure progress?#
Track content coverage, presence in answers, actual recommendations and downstream outcomes separately. A mention is not a recommendation, and a recommendation is not a sale.
Wide table: swipe horizontally or focus the table and use the arrow keys to see every column.
| Measure | Question | What it does not establish |
|---|---|---|
| Coverage | Is the verified fact present in the listing? | Presence in an AI answer |
| Presence | Does the brand or product appear in the recorded answer? | A recommendation to buy it |
| Recommendation | Is the product selected for the stated request? | A visit or purchase |
| Outcome | Did a measured referral, checkout or order follow? | Causality from the content change without an appropriate comparison |
Use a repeatable query set, geography, surface and observation method. Record the date and product version. Where feasible, test against a control. A before-and-after change in mentions alone does not isolate the effect of a catalog edit.
Merchant Center’s AI performance insights, open to US businesses since 16 September 2026, cover organic results only. A before-and-after reading needs a control to show what a change did. [A6]
How was the research conducted?#
Search universe and tags
598 US gift phrases: the most-searched “gifts for”, “christmas gift”, stocking stuffer, white elephant and secret santa phrases, plus 14 typed gift requests. Definitions, rules, wrapping, gift tags and DIY were left out. Gift bags, boxes, baskets and sets count as products. Multiple tags can apply to the same phrase. Search estimates were pulled on 20 October 2026.
AI Overview snapshots
The top 100 were ranked by December 2025 estimated US volume. The latest Ahrefs checks were dated 20 to 26 October 2026. Historical December snapshots were available for 94 phrases, dated 19 November to 14 December 2025. For the top 30 within that matched group, October 2025 snapshots were dated 18 to 19 October and the later 2025 snapshots were dated 25 November to 14 December.
A phrase counts as showing an AI Overview when Ahrefs recorded the feature on that result page. Results vary by user, device, geography and time. The report measures recorded presence, not every shopper’s exposure, impression counts or conversion.
Sensitivity and growth
“Gifts for kids” shows a large tracking change, from 23,234 estimated searches in December 2024 to 435,283 in December 2025. It remains in the December totals. Excluding it changes no reported intent share by more than 2.0 percentage points before rounding; the occasion share rises from 58.5% to 60.6%. Word-length growth uses only the 232 phrases tracked in 2019.
Outside evidence
Baymard’s figures come from a usability benchmark of 170+ sites and apps. The skincare study is a controlled preprint, not peer reviewed. The travel-bag example is Google’s own description of AI Mode. These sources should not be merged into one funnel or interpreted as a common sample.
Data availability
Lily AI analysis of Ahrefs Keywords Explorer US estimates, pulled 20 October 2026, and of Ahrefs’ records of Google’s US results dated 20 to 26 October 2026. Search volumes are Ahrefs’ modeled estimates, not Google’s counts. Raw Ahrefs exports are not published here. Independent studies are not Lily AI results.
Which 30 phrases are in the historical comparison?#
The volume column contains December 2025 estimated US searches, not AI Overview views. The three observation columns report whether an AI Overview was present in the recorded snapshot. The 30 are the biggest of the 94 phrases with a December 2025 snapshot; three phrases that name a year had none and would otherwise rank 19th, 27th and 29th. [A1]
Wide table: swipe horizontally or focus the table and use the arrow keys to see every column.
| Rank and phrase | Dec 2025 searches | Oct 2025 AI | Dec 2025 AI | Oct 2026 AI |
|---|---|---|---|---|
| 1. white elephant gift ideas | 964,875 | No | Yes | Yes |
| 2. gifts for mom | 556,168 | No | No | Yes |
| 3. gifts for dad | 463,044 | No | No | Yes |
| 4. gifts for kids | 435,283 | No | No | Yes |
| 5. white elephant gift | 431,819 | No | No | Yes |
| 6. gifts for men | 401,987 | No | No | Yes |
| 7. stocking stuffers | 322,667 | No | No | Yes |
| 8. christmas gifts | 266,941 | No | No | Yes |
| 9. christmas gift ideas | 257,398 | No | No | Yes |
| 10. gifts for women | 252,807 | No | No | Yes |
| 11. gifts for her | 221,260 | No | No | Yes |
| 12. white elephant gifts | 171,779 | No | No | Yes |
| 13. gifts for him | 160,174 | No | No | Yes |
| 14. best white elephant gifts | 131,076 | No | No | Yes |
| 15. christmas gift | 122,640 | No | Yes | Yes |
| 16. christmas gifts for women | 114,221 | No | No | Yes |
| 17. good white elephant gifts | 111,263 | No | No | Yes |
| 18. christmas gifts for mom | 104,797 | No | No | Yes |
| 19. funny white elephant gifts | 99,836 | No | No | Yes |
| 20. christmas gifts for men | 96,027 | No | Yes | Yes |
| 21. stocking stuffers women | 91,305 | No | Yes | Yes |
| 22. stocking stuffers for men | 89,993 | No | Yes | Yes |
| 23. mens stocking stuffers | 88,528 | No | Yes | Yes |
| 24. stocking stuffers men | 84,712 | No | Yes | Yes |
| 25. white elephant gift ideas $25 | 79,830 | No | No | Yes |
| 26. gifts for grandma | 65,908 | No | No | Yes |
| 27. secret santa gift ideas | 63,032 | No | Yes | Yes |
| 28. christmas gifts for dad | 58,224 | No | Yes | Yes |
| 29. secret santa gifts | 57,656 | No | Yes | Yes |
| 30. white elephant gift ideas $50 | 56,654 | No | No | Yes |
Sources#
- A1. Lily AI Research: Ahrefs US gift-search analysis
Original research dated 27 October 2026. Lily AI analysis of Ahrefs Keywords Explorer US estimates, pulled 20 October 2026, and of Ahrefs’ records of Google’s US results dated 20 to 26 October 2026. Raw exports are not published with this article.
- A2. Baymard Institute: Ecommerce Search UX 2026, search query types
Updated 29 April 2026. More than 170 sites and apps; more than 10,000 performance ratings. Independent benchmark.
- A3. Google Merchant Center: product data specification
Primary specification for the Google fields discussed, read with Google’s question_and_answer help page. Checked 29 October 2026. The comparison also draws on the Meta and Shopify specifications linked below.
- A4. Xi Chu and Yupeng Hou: Incumbent Advantage
arXiv 2606.17443v3, revised 10 October 2026. Baseline: experiment 1a, 670 valid trials. Interpolated thresholds: experiment 1c, 9,220 trials. Independent preprint, not peer reviewed.
- A5. Alphabet: Q3 2026 earnings release, Exhibit 99.1 to Form 8-K
28 October 2026. $56,567 million in Q3 2025 and $63,350 million in Q3 2026. “Search and other” is broader than search ads alone.
- A6. Google on AI Mode and shopping
AI Mode shopping announcement, 20 May 2025; agentic checkout, 13 November 2025; Sundar Pichai at NRF, 11 January 2026; Alphabet earnings remarks, 22 July and 28 October 2026; Merchant Center holiday updates, 16 September 2026; Universal Cart, 15 October 2026. The link opens the NRF remarks.
- A7. Shopify: open product taxonomy
Primary taxonomy repository. Release 2026-10, checked 29 October 2026.
- A8. Meta: catalog field reference
Primary documentation reference, checked 29 October 2026. Platform documentation can change.
Where Lily AI fits#
Lily Max is the agentic product intelligence engine for AI commerce. It enriches retail product catalogs and measures results against a control across Google Ads, Meta Ads, AI shopping assistants and onsite search.
A free feed audit looks at your Google Merchant Center feed and shows where richer product data can lift Shopping and AI-discovery performance.
Get a free feed auditThe complete report is freely readable. You do not need the feed audit to read any part of this report.