2026 AI SEARCH DATA RESEARCH · GEO

GEO Research on AI Marketing Creative Tools: Similar AI mention rates, but nearly twice the chance of being recommended first (139,855 AI answers)

Article contents

Across 31 days, seven AI platforms, and 139,855 answers about AI marketing creative tools, brands with similar AI mention rates can have nearly twice the chance of being the first recommendation.

Author Dageno Research TeamPublished November 2026Data Dageno AI Answers Database, Oct 17–Nov 16, 2026Sample 139,855 answers · 7 AI platforms

Key findings

Short answer: Brand mention rate is only the first metric to track in GEO. In core questions about AI marketing creative tools, Google and Creatify appear at nearly identical rates (24.24% vs. 23.65%), yet when mentioned, Google is recommended first in just 17.8% of cases versus 35.0% for Creatify—almost twice as often. The leading brand also changes by question type and AI platform.
Figure 1. Four key GEO metrics
Figure 1. Mentions, first recommendations, and citations are different measures. Seven AI platforms, October 17–November 16, 2026. Panel 01 is calculated from report values; panel 02 reflects reported chart values.
  1. 17.8% vs. 35.0%: In the core question set, Google and Creatify have similar mention rates (24.24% and 23.65%). When mentioned, Google ranks first 17.8% of the time, versus 35.0% for Creatify.
  2. 69.7% vs. 56.2%: Meta appears in 69.7% of answers about ad testing and optimization. For UGC and virtual spokespersons, Creatify leads with 56.2%. A single product-category average would hide the difference.
  3. 25.6% vs. 70.7%: Gemini includes web citations in 25.6% of its answers; Google AI Mode does so in 70.7%. Their brand-naming rates are similar, however (90.8% and 88.3%).
  4. 69.2% vs. 4.5%: Of cited sources, counted per answer × domain, 69.2% are brand or company websites; community, video, and review sites account for only 4.5%.

For brands making AI-generated video or ad creative, the metric most often monitored in GEO is the percentage of AI answers that mention the brand, called the mention rate (or “appearance rate” in the underlying database). GEO providers serving this industry often build query sets around broad product categories such as “AI video generation” and “AI image generation,” rank brands by mention rate, and use that ranking in proposals and KPIs.

But two brands mentioned almost equally often may have very different chances of being the first recommendation. We examined 139,855 answers collected over 31 days in Dageno’s AI Answers Database to see what gets missed when looking only at mention rates. The data covers seven AI platforms and records which brands each answer names, which brand is named first, and which web pages are cited.

Three metrics used in this article: Mention rate is the percentage of answers that name a brand. First pick is the first brand recommended in an answer. Conditional first-pick rate is the percentage of brand mentions in which that brand appears first—first-pick count divided by mention count. These metrics describe AI answers; they do not assess product quality or establish causality.

01 | When brands are mentioned equally often, how different is their chance of ranking first?

Almost twice as often. Platforms are mentioned frequently but less often rank first when mentioned; specialist marketing-creative tools show the reverse pattern.

We divided each brand’s first-pick count by its mention count in the core question set (46,309 answers, of which 89.6% named a brand). The table shows the eight brands appearing in both the top 12 by mention rate and top nine by first-pick share.

BrandMention rateFirst-pick shareConditional first-pick rate
Creatify23.65%9.25%35.0%
AdCreative.ai19.04%7.14%33.6%
Meta25.79%8.48%29.5%
Canva20.67%6.37%27.6%
OpenAI8.87%2.60%26.3%
Adobe16.56%4.14%22.4%
HeyGen21.35%4.77%20.0%
Google24.24%4.82%17.8%

Mention rate uses all 46,309 answers as its denominator. First-pick share uses the 41,496 answers naming at least one brand. The first two columns come from report charts; the last is calculated.

Figure 2. Brand mention rate versus conditional first pick
Figure 2. A high mention rate does not necessarily mean a high conditional first-pick rate. Core question set, October 17–November 16, 2026. Orange indicates specialist tools; blue indicates broader platforms (author classification).

Among these eight brands, Google ranks second in mention rate but last in conditional first-pick rate. Not every specialist tool performs strongly either: HeyGen’s conditional first-pick rate is only 20.0%.

Another example outside the table is Photoroom. Its first-pick share is 3.59%, while its mention rate is no more than 7.91% (outside the top 12). By the same formula, its conditional first-pick rate is approximately 40.7% or higher—above Creatify.

Why do platform companies rank first less often when mentioned? They may be included as supporting alternatives rather than the primary answer. This is an inference: verifying it would require examining the position and context of each mention, which this dataset cannot do.

For brands: Mention rate and conditional first-pick rate answer different questions. A low mention rate means AI is leaving you out of the shortlist. A reasonable mention rate with a low conditional first-pick rate means you make the list but rank farther down. Those gaps call for different investigations. For agencies: Include both metrics in reports and KPIs. Mention rate alone can overstate Google-like companies’ ranking advantage and understate specialist tools such as Creatify.

02 | Can broad product categories point to the wrong competitors?

Quite possibly. Each of the three broad product categories has a different set of leading brands.

Product categoryAnswersTop three by mention rate
AI video generation73,922Kling AI 36.2%, Google 30.1%, HeyGen 20.0%
AI image generation38,793Adobe 33.5%, OpenAI 31.6%, Midjourney 31.5%
AI ad creative generation27,140Meta 45.0%, Creatify 36.5%, AdCreative.ai 32.6%

An additional 13.9% of answers name no brand at all. First place is highly fragmented: overall, even Google—the brand most frequently recommended first—accounts for only 9.7%, followed by 7.2% for the runner-up. This does not contradict Section 01: because Google appears often, it collects many first picks in absolute terms; when normalized by mentions, its first-pick rate is the lowest of the eight brands.

The leading brands in AI video generation and AI ad creative generation barely overlap. An ecommerce ad-video tool looking only at the broader video category might identify Kling AI as its main competitor, while the leading names in ad-creative queries are Meta, Creatify, and AdCreative.ai.

For agencies: Do not define the competitor set solely by the client’s self-described product category. If a client spans two categories, build a separate competitor list for each; otherwise, a proposal may begin with the wrong competitors.

03 | Does the question topic change which brand leads?

To a large extent. Specialist tools lead many asset-creation questions, while platforms lead campaign management, procurement, and workflows.

Figure 3. Leading brands across 11 question topics
Figure 3. Specialist tools lead many content-creation tasks; platform companies lead many media-buying and workflow tasks. Eleven question topics, October 17–November 16, 2026.
Question categoryQuestion topicsLeading brands
Creating assetsUGC & virtual spokespersons; video ads; product videos; product imagery & photography; editing & repurposing; social short videosCreatify, HeyGen, Photoroom, OpusClip, CapCut
Managing campaigns & workflowsAd testing & optimization; batch creation & workflows; pricing & procurement; brand consistencyMeta, Google, Adobe
ExceptionAd visual designMeta 46.5%; AdCreative.ai 42.9%

“Ad visual design” sounds like a creative-production task, but Meta leads, with AdCreative.ai close behind. We therefore list it separately as an exception.

Concentration varies markedly by topic. “Ad testing and optimization” is the most concentrated: Meta appears in 69.7% of answers, followed by AdCreative.ai at 54.9%. Four topics have leaders below 30%: pricing and procurement (28.6%), social short videos (26.7%), batch production and workflows (26.1%), and video editing and content repurposing (24.8%; only 588 answers). The top three are within five percentage points in three of these topics: video editing and repurposing (24.8%, 22.3%, 21.3%), pricing and procurement (28.6%, 26.2%, 24.3%), and social short videos (26.7%, 25.1%, 22.1%).

For brands: First identify which tasks you underperform in. For a UGC video tool, its position in “UGC and virtual spokespersons” is more revealing than its average mention rate across all video-generation queries. For agencies: Split question sets by buyer task and report a separate baseline for each. Concentrated topics such as ad testing and diffuse topics such as social short videos should not share one target: in the former, first check whether the brand makes the list; in the latter, track movement in ranking.

04 | Do rankings and citations change across AI platforms?

No. Brand-naming rates are relatively similar (79.9%–96.2%), but web-citation rates differ much more.

PlatformAnswersBrand namedWeb page citedTop three by mention rate
Perplexity52,77179.9%26.3%Google 15.7%, Kling AI 13.3%, HeyGen 11.0%
ChatGPT35,80588.9%33.8%Google 29.9%, Adobe 28.0%, Kling AI 22.3%
Gemini34,33690.8%25.6%Google 31.9%, Kling AI 23.7%, Adobe 23.1%
Google AI Mode9,86388.3%70.7%Google 45.8%, Kling AI 31.9%, Synthesia 24.6%
Google AI Overviews4,18596.2%61.2%Google 34.6%, Kling AI 31.7%, Meta 18.7%
Copilot2,05687.3%23.5%Google 34.0%, OpenAI 28.6%, Adobe 25.7%
Grok77790.9%32.2%Meta 42.3%, Google 39.0%, AdCreative.ai 26.6%

Answer counts describe platform sample sizes only. Grok and Copilot have relatively few answers, so their results are directional. Question mixes are not identical across platforms.

Figure 4. Brand mentions and citations across AI platforms
Figure 4. All platforms name brands, but the frequency of web citations varies widely. October 17–November 16, 2026.

Google ranks first on six of seven platforms; Grok is the exception. It is highest on Google AI Mode (45.8%) but also more prominent in Copilot (34.0%) than in Gemini (31.9%), so it is not enough to say Google simply favors its own brand on its own platform.

Perplexity has the lowest brand-naming rate and the most dispersed shortlist; its leading brand appears in only 15.7% of answers. Just 26.3% of its answers contain web citations, contrary to the common impression that Perplexity always provides sources. This may reflect how this database detects citations, so we do not draw a firm conclusion.

For agencies: The share of answers with web citations differs by roughly threefold across platforms (70.7% on Google AI Mode versus 23.5% on Copilot). On Google AI Mode and AI Overviews, roughly 60–70% of answers cite pages, so citation analysis covers more responses. On Gemini and Copilot, only about one-quarter do; brand-naming data is the more useful way to read those shortlists. Platform-level reporting is more informative than one cross-platform average.

05 | What kinds of web pages do AI answers cite?

Mostly brand and company websites; communities and review sites contribute little.

Figure 5. Types and domains of AI citation sources
Figure 5. Nearly seven in ten cited sources are brand or company websites. Counted per answer × domain, October 17–November 16, 2026. Site categories are author classifications.

Across the complete dataset, 32.2% of answers cite a web page. Counted by answer × domain, brand and company websites account for 69.2% of cited sources, media/directories/other sites for 26.3%, and community/video/review sites for 4.5%. Of the twelve most-cited domains, nine are tool brand sites; the other three are YouTube, Reddit, and Zapier. The leading domains, canva.com and heygen.com, each account for only 2.45% of citations, suggesting a dispersed citation landscape.

Two details stand out. higgsfield.ai ranks fourth by citation share (1.93%), yet Higgsfield does not appear in the top nine brands by mention rate across the three broad product categories and enters the top three only in South Korea among the twelve reported markets. Its site is cited often while its brand is named relatively rarely. Also, runway.com and runwayml.com appear as separate entries: two domains belonging to the same brand split its citation share.

The dominance of official websites is consistent with Dageno’s earlier analysis of 10.05 million answers, Which websites do AI answers cite most often?, where brand sites were likewise the largest source category. The separation between citations and brand mentions is directionally consistent with When AI cites an external page, does it mention the brand? The studies cover different platforms, industries and periods, so only their direction—not their percentages—can be compared.

For brands: Official sites make up the largest source category in this industry. Check whether your own site is cited, which pages are cited, and whether citations coincide with brand mentions. For agencies: Merge multiple domains belonging to the same brand when reporting citations. A brand like Higgsfield, which receives citations but relatively few mentions, may look more competitive than it is if citations are viewed in isolation.

06 | Do AI platforms recommend the same brands in different markets?

Not necessarily. Germany and the US share the same top three, while India and Japan differ markedly. Most non-US market samples are small, so treat them as directional.

MarketAnswersTop three by mention rate
United States127,199Google 26.8%, Kling AI 20.8%, Adobe 19.2%
United Kingdom2,641Adobe 25.8%, Canva 17.1%, Google 15.6%
Germany1,917Google 24.8%, Kling AI 19.0%, Adobe 18.0%
India1,347Meta 65.0%, AdCreative.ai 41.4%, Google 32.8%
Japan320Meshy 32.5%, Tripo AI 27.2%, Adobe 14.7%

Answer counts indicate sample size, not market size.

Germany has the same top three as the United States, in the same order; Adobe leads in the UK. India strongly skews toward ad-buying tools, with Meta appearing in 65.0% of answers. Japan’s top two, Meshy and Tripo AI, mainly generate 3D models. Differences may be driven by differing question mixes across markets; raw prompts would need to be sampled to confirm.

For agencies: In multi-market projects, do not treat the US shortlist as a global shortlist. Check the top three in each market before deciding whether to localize the question set. Markets with only a few hundred answers (such as Japan’s 320) are useful signals, not a sound basis for conclusions or KPIs.

07 | How can brands and agencies use these GEO findings?

Taken together, the findings suggest a four-layer approach to diagnosing a brand’s position. The framework cannot guarantee recommendations; it helps locate where the shortfall occurs.

LayerQuestion to answerWhat to measure
1. Task layerWhich task is the buyer asking about?Task-based question topics and answer counts
2. Shortlist layerDoes my brand make the shortlist?Mention rate by topic and platform
3. Ranking layerWhere do I rank when mentioned?Conditional first-pick rate versus key competitors
4. Source layerWhich pages is AI citing?Source categories, citations to owned domains, and web-citation rates by platform
  1. Build question sets by task, separating “creating assets” from “managing campaigns” at minimum. In the same market, Meta leads ad testing and optimization (69.7%), while Creatify leads UGC and virtual spokespersons (56.2%). A single average masks the distinction (Section 03).
  2. Report mention rate and conditional first-pick rate together. In the core question set, Google and Creatify have similar mention rates, but their conditional first-pick rates differ by nearly 2× (17.8% versus 35.0%). The follow-up investigations are different (Section 01).
  3. Separate citation data by platform and consolidate multiple domains for the same brand. The share of answers with citations ranges from 23.5% to 70.7%; runway.com and runwayml.com being counted separately illustrates the domain issue (Sections 04–05).
  4. Check sample counts for each market before incorporating a result into a KPI. India and Japan differ sharply from the United States, but have just 1,347 and 320 answers, respectively (Section 06).
  5. Do not judge from one-off results. In another Dageno study, repeating the same question later that day (June–September 2026; five AI platforms; 7,030 comparisons), ChatGPT returned exactly the same brand list only 7.4% of the time after excluding pairs in which neither answer named a brand. Evaluate proportions over a period of time, not one screenshot.

Want these numbers for your own industry? Public market data is available free after registration on Dageno, covering 10 AI platforms, 86 countries and regions, and 12,000+ niche markets. Agencies can explore the Dageno Partner Program to use task-level breakdowns in proposals and client reviews. White-label agency reports are coming soon.

Research methodology

Data period31 daysOct 17–Nov 16, 2026
Answer count139,855Core question set: 46,309
AI platforms7Perplexity, ChatGPT, Gemini, Google AI Mode, Google AI Overviews, Copilot, Grok
Query set3 categories · 11 topicsAI video, AI image, AI ad creative

Definitions: Mention rate = answers naming the brand / all answers. First pick = the first recommended brand in an answer; first-pick share uses answers naming any brand as denominator. Conditional first-pick rate = first-pick count / mention count, calculated in this article as “first-pick share × 41,496 ÷ (mention rate × 46,309).” Source types are counted per answer × domain. Specialist-tool versus platform classifications are editorial judgments.

Sample composition and limitations: Answers cover 54 countries and regions, but the United States accounts for 91.0% of responses. Overall figures therefore mostly describe the US sample. This reflects the database’s collection mix, not market share. Grok, Copilot, and most non-US markets have smaller samples; questions vary across platforms and markets. Findings are observational associations, not causal evidence.

References

Dageno AI Answers Database, October 17–November 16, 2026.
Dageno Research Team, “When AI cites an external web page, does it name a brand?” https://www.dageno.ai/zh/research/ai-citations-vs-brand-mentions
Dageno Research Team, “Which websites do AI answers cite most? ChatGPT and Google cite different sites.” https://www.dageno.ai/zh/research/ai-most-cited-websites-chatgpt-vs-google
Dageno Research Team, “Does asking AI the same question again on the same day return the same recommended brands?” https://www.dageno.ai/zh/research/ai-same-day-repeat-queries

DAGENO RESEARCH