- Home
- Insights
- Growth Marketing
- How does your brand appear in AI answers? A step-by-step guide to GEO and AEO
How does your brand appear in AI answers? A step-by-step guide to GEO and AEO
Some customers no longer look at search results but at an assistant’s answer. This guide covers the technical conditions for appearing in that answer, how to write content for it and how to measure visibility.
What do GEO and AEO change?
In classic search the user sees a list and clicks. An assistant answers the question itself and shows a few sources under the answer. The brand’s aim is less to rank high in a list and more to have its name or its page appear in that answer.
GEO (Generative Engine Optimization) is the work done so that content appears as a source in generative AI answers. AEO (Answer Engine Optimization) is structuring content so that systems which answer a question directly select it as the answer. Both build on good SEO; neither replaces it.
What do Google’s AI features require?
According to Google, appearing in AI Overviews or AI Mode needs no additional requirement and no special optimisation. Per Google Search Central’s guidance, for a page to be eligible as a supporting link in these features it must be indexed and eligible to be shown in Google Search with a snippet; there are no further technical requirements.
The same page makes three more points:
- No new machine-readable files, AI text files or special schema.org markup are needed for these features.
- Traffic from these features is included in overall search traffic in Search Console’s Performance report, under the “Web” search type.
- Site owners who want to limit what is shown can use the
nosnippet,data-nosnippet,max-snippetandnoindexcontrols.
How do ChatGPT and Claude read your site?
Both companies use separate crawlers for separate jobs, and each can be allowed separately in robots.txt. The table below is compiled from the companies’ own documentation.
| Crawler | What for | If blocked |
|---|---|---|
| OAI-SearchBot | Surfacing websites in ChatGPT’s search features. | The site is not shown in ChatGPT search answers; it can still appear as a navigational link. |
| GPTBot | Crawling content that may be used to train OpenAI’s models. | Signals that the content should not be used in training. |
| ChatGPT-User | Opening a page at a user’s request; it does not crawl automatically. | OpenAI notes that robots.txt rules may not apply because the request is user-initiated. |
| Claude-SearchBot | Improving the quality of Claude’s search results. | The site’s visibility in user-directed search may be reduced. |
| ClaudeBot | Collecting content that could contribute to model training. | Signals that the site’s future content should be excluded from training data. |
| Claude-User | Opening a page when a user asks Claude a question. | Claude cannot open your page when answering a user’s request. |
OpenAI states plainly that the settings are independent: a site can allow OAI-SearchBot to appear in search results while disallowing GPTBot for training. Anthropic also describes its three crawlers separately. You can prepare which crawler your site allows with the AI crawler robots.txt tool.
Is llms.txt necessary?
It is not required. llms.txt is a proposal Jeremy Howard published on 3 September 2024: a markdown file at the site root that summarises for AI systems what the site is and which pages matter.
Google says no such file is needed for its own AI features. In our view llms.txt is harmless and a small job, but it does not replace robots.txt or well-written pages. This site also has an llms.txt file.
How is content written so it can be cited?
For an assistant to cite a sentence, that sentence has to make sense on its own and be verifiable. The structure below is also used in the articles on this site; it is not a guarantee, but a way of writing that makes content easier to read.
- A short answer under the title. The answer to the question is given in 40–60 words at the top of the page.
- Headings phrased as questions. Written the way people ask an assistant; the first sentence under the heading is the answer.
- One-sentence definitions. A term is defined once where it first appears and then always used under the same name.
- Author and date visible. Who wrote the page and when it was updated is stated on it.
- Linked sources. Every figure and claim links to its source; a figure without a source is not written.
How is visibility measured?
In two ways. The first is Search Console: traffic from Google’s AI features appears inside total web traffic in the Performance report. It does not show as a separate line, so the change at page level is what is watched.
The second is a question panel. Write down 10–20 questions your customers might ask an assistant and ask the same questions of ChatGPT, Gemini, Claude and Perplexity every month. For each answer, record two things: is your brand named, and is your site linked? Because assistants’ answers can vary from question to question and day to day, the trend over months matters, not a single try.
In your analytics tool you can also track referrals from AI assistants’ domains as a separate group.
Our recommendation
Start not by writing content but by checking access. Content written for a site that cannot be read is wasted.
- Check that your robots.txt does not block OAI-SearchBot and Claude-SearchBot. A firewall or bot protection may block these crawlers separately; look at the server logs.
- Make a separate decision about training crawlers. You can block or allow GPTBot and ClaudeBot without affecting search visibility.
- Confirm in Search Console that your ten most important pages are indexed.
- List the questions your customers ask most often and write a page with a short answer for each.
- Set up the question panel, take the first measurement today and repeat it every month. After three months, decide from the trend which pages to strengthen.
Frequently asked questions
Is special markup or a special file needed to appear in AI answers?
According to Google, not for its own AI features; it asks for no machine-readable files, AI text files or special schema.org markup. For other assistants too, what decides it is that the page can be crawled and understood.
If I block training crawlers, will I disappear from search answers too?
According to OpenAI, no; the settings are independent. GPTBot can be blocked while OAI-SearchBot is allowed. At Anthropic too, ClaudeBot for training and Claude-SearchBot for search are managed separately.
When will the results of changes show?
There is no fixed time. Google states that recrawling after changes can take anywhere from several days to several months. That is why measurement is monthly and decisions are not made on weekly swings.
Sources
- Google Search Central: AI features and your websiteAccessed 19 November 2026
- OpenAI: Overview of OpenAI CrawlersAccessed 19 November 2026
- Anthropic: Does Anthropic crawl data from the web, and how can site owners block the crawler?Accessed 19 November 2026
- llmstxt.org: The /llms.txt file (Jeremy Howard)3 September 2024
Read next
All insightsHow is the learning phase budget set for Meta ads?
Ad sets launched on Meta with a low budget often get stuck in the learning phase. This guide covers how to calculate the minimum budget, what to do when the budget falls short and what resets learning.
Read the article Growth MarketingWhat should ROAS be? Guide · 6 min readWhat is break-even ROAS and how is it calculated?
A high ROAS does not show that advertising leaves a profit. The point where profit begins is set by gross margin. This guide covers break-even ROAS, target ROAS and the most common mistakes in the calculation.
Read the article