# New brands: why AI doesn't mention you yet, and what to do first

> AI answers favour brands the web already writes about, and a model may know nothing about you. Make your site reachable and unambiguous first, publish one clear facts page, then win narrow questions and your first independent coverage. Measure with enough repeats, because small brands appear unevenly.

- URL: https://geo.casa/best-practices/new-brand/
- Guide · Officially documented
- Reviewed: 2026-10-28

This guide is for founders and marketers at a brand that is new, small, or known only in a niche. It explains why AI assistants don't name you yet and gives the first six things to do, in order. Most of the early steps take days, not months. Earning outside coverage is the slow part, so start it early.

## The short version

1. Ask the assistants what they know about you, and accept that the answer may be nothing.
2. Make sure search engines and AI crawlers can reach and index your site.
3. Publish one clear facts page and mark up your organisation, so you can't be confused with anyone else.
4. Use the official listings that apply to you: business profiles, product feeds, app stores.
5. Aim at narrow questions where you are a real answer, not the category's head question.
6. Earn your first independent coverage and reviews.
7. Measure with enough repeats to see a small brand's uneven appearances.

## Why AI doesn't mention you yet

Two things work against a new brand. Models learned from text written before you existed. And when they search, they lean on sources that already write about established names.

> **Officially documented:**
> OpenAI says its models are trained on data up to a cutoff and don't know about later events unless tools such as search are used. Without search, ChatGPT's answers rest on what the model learned in training. With search, it can cite current web sources.
>
> Sources: [OpenAI Help Center: Does ChatGPT tell the truth?](https://help.openai.com/en/articles/8313428-does-chatgpt-tell-the-truth)

> **Research-backed:**
> A 2025 University of Toronto study asked ChatGPT and Perplexity 50 unbranded cola questions that asked for the most popular, best or top brands. Major brands took 56.3% of ChatGPT's brand mentions and niche brands 12.3%. On Perplexity the split was 67.9% major and 5.8% niche. The authors call this a "big brand bias" that niche players must overcome.
>
> Sources: [Chen, Wang, Chen and Koudas (University of Toronto), arXiv:2509.08919: Generative Engine Optimization: How to Dominate AI Search](https://arxiv.org/abs/2509.08919)

> **Research-backed:**
> An EMNLP 2024 study across four product categories found that language models disproportionately associated global brands with positive attributes, compared with local brands.
>
> Sources: [Kamruzzaman, Nguyen and Kim, arXiv:2406.13997 (EMNLP 2024): "Global is Good, Local is Bad?": Understanding Brand Bias in LLMs](https://arxiv.org/abs/2406.13997)

> **Observed in practice:**
> In DATA17's August 2026 sample of 72 prompts in one software category, only 18 of the category's 171 brands were mentioned at all.
>
> Sources: DATA17 AI Research Lab · 解数智能: 你的公司还在 AI 的答案里吗：企业 GEO 行动指南 · 2026 (unpublished report, not publicly available; readers can't check it)

The cola questions asked for "popular" brands, which favours the leaders by design. Still, the pattern is consistent: in a broad category question, most brands never appear. That is the normal starting point, not a penalty.

> Not being named is the default for a new brand. It tells you what to build, not that something is broken.

## Step 1: find out what assistants know

Ask the assistants your customers use some direct questions: "What is [brand]?", "Who makes [product]?", "Is [brand] any good?". Ask with search on and off where you can. Record whether they know you, confuse you with someone else, or invent details. If your name is shared with another company or a common word, note it now. It shapes step 3.

## Step 2: be reachable and indexed

AI search can only cite pages its search systems can fetch. For a new site this is often where things fail: the site isn't indexed yet, or a firewall blocks crawlers.

> **Officially documented:**
> To appear as 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".
>
> Sources: [Google Search Central: AI features and your website](https://developers.google.com/search/docs/appearance/ai-features)

> **Officially documented:**
> OpenAI: "Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though can still appear as navigational links."
>
> Sources: [OpenAI: Overview of OpenAI crawlers](https://developers.openai.com/api/docs/bots)

Run the [AI search access audit](https://geo.casa/playbooks/ai-search-access-audit/) on your homepage, product pages and facts page. Submit a sitemap to Google and Bing, and to Baidu if you sell in China ([GEO in China: Doubao, Yuanbao, Qwen, ERNIE, DeepSeek and Kimi](https://geo.casa/best-practices/china-platforms/)).

## Step 3: make yourself unambiguous

Search engines build entity records from many sources. A new brand has few sources, so the ones you control need to be clear and consistent.

> **Officially documented:**
> Google says "Knowledge Panels are created automatically by Google Search Algorithm when there is enough information available on the open web." Facts come from a variety of sources, including public sources, licensed data and content owners who suggest changes to panels they have claimed.
>
> Sources: [Google Knowledge Panel Help: How Google's Knowledge Graph works](https://support.google.com/knowledgepanel/answer/9787176?hl=en)

> **Officially documented:**
> "Adding organization structured data to your home page can help Google better understand your organization's administrative details and disambiguate your organization in search results." The `sameAs` property lists your profile pages on other sites, such as social media or review sites.
>
> Sources: [Google Search Central: Organization schema markup](https://developers.google.com/search/docs/appearance/structured-data/organization)

Do this:

- **One facts page** (often the About page): legal name, brand name, what you sell, who it's for, founding year, location, founders, and links to official profiles. Write it in plain sentences an assistant could quote.
- **Organization structured data** on the homepage or facts page, with `sameAs` pointing to your real profiles ([Organization structured data](https://geo.casa/content-structured-data/organization-structured-data/)).
- **The same name and description everywhere**: social profiles, marketplaces, directories and app stores.

## Step 4: use the official listings that apply

For local businesses and products, platforms take structured data directly. This is the fastest route for a new brand to be described correctly.

> **Officially documented:**
> Google says generative AI responses can include product listings, product information and local business information. Merchant Center feeds and Google Business Profiles can help products and services appear in AI responses and other search results.
>
> Sources: [Google Search Central: Optimizing your website for generative AI features on Google Search](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide)

If you have a physical location or service area, set up a Business Profile. If you sell products online, submit a product feed. For apps, keep store listings complete. In China, the equivalent is to claim your listings on the map, review and marketplace platforms your customers use.

## Step 5: aim at narrow questions first

A new brand rarely wins "best running shoes". It can win "running shoes for wide feet under $100" if that is truly its strength. Choose 10–20 specific questions where you are a genuine answer. Then publish a page that answers each one directly, with conditions and limits ([For content editors: writing articles AI answers can quote](https://geo.casa/best-practices/content-editor/)).

> Win the question you are really the best answer to. Broad category questions come later, if at all.

## Step 6: earn your first independent coverage

Assistants and encyclopedias both look for what others say about you. A new brand's first goal is a handful of genuine third-party sources: a trade-press story, a review, an inclusion in a relevant comparison, a customer case published by the customer. How to earn them is in [PR and brand teams: how press, encyclopedias, Q&A sites and reviews shape AI answers](https://geo.casa/best-practices/pr-earned-media/).

> **Hypothesis:**
> Meltwater's 2026 report argues that when there is little user-generated discussion of a brand, AI tends to treat it as a low-activity or low-relevance option. The report gives no test of this.
>
> Sources: Meltwater 融文: 2026中国企业出海GEO洞察报告——AI时代的品牌信任重构 (unpublished report, not publicly available; readers can't check it)

Don't start with Wikipedia. Its notability rules require substantial coverage in independent sources, which a new brand usually lacks ([PR and brand teams: how press, encyclopedias, Q&A sites and reviews shape AI answers](https://geo.casa/best-practices/pr-earned-media/)).

## What not to do

- **Buy a burst of mentions.** Google treats it as spam (below), and seeded content is the target of China's 2026 campaign ([Claim: seeding brand mentions across the web boosts AI visibility](https://geo.casa/ai-search/inauthentic-mentions/), [GEO regulation in China](https://geo.casa/foundations/geo-regulation-china/)).
- **Publish dozens of thin pages.** Scaled low-value content is a spam pattern ([Scaled content abuse](https://geo.casa/content-structured-data/scaled-content-abuse/)).
- **Pay for a guaranteed "AI ranking".** No documented method guarantees it ([Before you hire a GEO agency](https://geo.casa/best-practices/choose-agency/)).
- **Judge from one screenshot.** Small brands appear unevenly. One run proves nothing either way.

> **Officially documented:**
> Google's definition of spam includes "attempting to manipulate generative AI responses in Google Search."
>
> Sources: [Google Search Central: Spam policies for Google web search](https://developers.google.com/search/docs/essentials/spam-policies)

## How you'll know it's working

> **Observed in practice:**
> In a September 2025 study across three Chinese models, Miaozhen found that results reached basic stability only at 30 or more runs per prompt. Brands with low visibility (around 20%) needed 60 runs or more.
>
> Sources: [中国商务广告协会AI营销应用工作委员会 · 秒针营销科学院: 2025年中国GEO行业发展报告](https://g4-img-all.oss-cn-beijing.aliyuncs.com/video/50108560/01503a0d1eaeb4429ee3c0014738e9ea.pdf)

- **Recognition first.** Do assistants now describe you correctly when asked by name? This usually moves before category mentions.
- **Narrow questions next.** Track your 10–20 target questions with enough repeats ([Measuring AI visibility by sampling prompts](https://geo.casa/measurement/ai-visibility-measurement/)).
- **Cited sources.** Are your facts page and your new third-party coverage being cited?
- **Search Console and Bing.** Are the pages indexed and getting impressions ([Generative AI performance report: AI Overviews and AI Mode in Search Console](https://geo.casa/measurement/search-console-ai-features/), [Bing AI Performance report](https://geo.casa/measurement/bing-ai-performance/))?

## Evidence at a glance

| Recommendation | Evidence |
|---|---|
| New brands start out unmentioned in broad questions | Research and observation |
| Models don't know about you without search | Official (OpenAI) |
| Be indexed and crawlable first | Official |
| Facts page plus organization markup for disambiguation | Official (Google) |
| Use business profiles and product feeds | Official (Google) |
| Start with narrow questions | Our recommendation |
| Thin user discussion keeps a brand out of answers | Hypothesis (vendor report, untested) |
| Measure low-visibility brands with 60+ runs | Observed |

## Sources

- [Chen, Wang, Chen and Koudas (University of Toronto), arXiv:2509.08919: Generative Engine Optimization: How to Dominate AI Search](https://arxiv.org/abs/2509.08919)
- [Kamruzzaman, Nguyen and Kim, arXiv:2406.13997 (EMNLP 2024): "Global is Good, Local is Bad?": Understanding Brand Bias in LLMs](https://arxiv.org/abs/2406.13997)
- DATA17 AI Research Lab · 解数智能: 你的公司还在 AI 的答案里吗：企业 GEO 行动指南 · 2026 (unpublished report, not publicly available; readers can't check it)
- [OpenAI Help Center: Does ChatGPT tell the truth?](https://help.openai.com/en/articles/8313428-does-chatgpt-tell-the-truth)
- [Google Knowledge Panel Help: How Google's Knowledge Graph works](https://support.google.com/knowledgepanel/answer/9787176?hl=en)
- [Google Search Central: AI features and your website](https://developers.google.com/search/docs/appearance/ai-features)
- [OpenAI: Overview of OpenAI crawlers](https://developers.openai.com/api/docs/bots)
- [Google Search Central: Organization schema markup](https://developers.google.com/search/docs/appearance/structured-data/organization)
- [Google Search Central: Optimizing your website for generative AI features on Google Search](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide)
- Meltwater 融文: 2026中国企业出海GEO洞察报告——AI时代的品牌信任重构 (unpublished report, not publicly available; readers can't check it)
- [中国商务广告协会AI营销应用工作委员会 · 秒针营销科学院: 2025年中国GEO行业发展报告](https://g4-img-all.oss-cn-beijing.aliyuncs.com/video/50108560/01503a0d1eaeb4429ee3c0014738e9ea.pdf)
- [Google Search Central: Spam policies for Google web search](https://developers.google.com/search/docs/essentials/spam-policies)