# Do you need a GEO agency? What they do, and what you can do yourself

You want your product named in AI answers. You do not want to pay for fluff. This guide shows what a GEO agency really does, what you can ship in an afternoon, what to ask before you sign, and when the spend is worth it.

Updated 2026-09-14 · Source: https://porteur.ai/guides/do-you-need-a-geo-agency

## What a GEO agency actually does

GEO is getting cited in answers from ChatGPT, Perplexity, Google AI Overviews and AI Mode, Claude, Gemini and Copilot. Agencies package four things.

- Monitoring: they ask buyer questions on assistants, record who gets cited, and track share of mentions over time.
- Content: they write pages that answer questions directly with clear headings, lists and tables, and add dates and authors.
- Listings outreach: they research roundups and comparison pages assistants already cite, then ask editors to add your product with a fair summary.
- Technical access: they fix robots.txt to allow the answering crawlers, make content visible without JavaScript, and clean up site structures that confuse indexes.

Some add reporting and training. Some bundle classic SEO: keyword research, internal links and schema. That can help GEO, but it is not the same service.

## What is real and what is repackaged SEO

Real GEO work matches how assistants pick sources as of 2026: retrieve from a web index, cite pages that state the answer early, and prefer consistent names, lists and figures they can quote.

- Real: letting OAI-SearchBot, Claude-SearchBot, PerplexityBot, Bingbot and Googlebot crawl your site, while keeping training bots optional.
- Real: writing pages that start with the answer in the first paragraph, with headings that match the question, and a recent visible date and author.
- Real: getting on comparison pages and roundups that assistants already cite for your topic.
- Real: measuring citations by asking assistants your buyer questions on a schedule and logging who they name.

- Repackaged: selling “GEO keywords” lists without checks in assistants.
- Repackaged: only doing schema or meta descriptions and calling it GEO.
- Repackaged: pitching paid placements inside AI answers. You cannot buy citations.
- Repackaged: blocking or ignoring answering crawlers, then promising visibility in assistants.

## The part you can do this afternoon

1. **Open crawler access without training your content by default** Edit robots.txt to allow the answering and search crawlers. Keep training crawlers optional. For OpenAI, allowing OAI-SearchBot while disallowing GPTBot keeps you in ChatGPT search results and out of model training.
2. **Make one quotable page pattern** Pick a buyer question for a core page. Example: "/guides/getting-started" or "/pricing" or "/alternatives/competitor". Start with a two sentence answer, add a numbered list or table, a date, and an author line. Use consistent product and plan names.
3. **Get on the lists assistants already cite** In Perplexity and ChatGPT search, ask: "best [your niche] tools", "[your product] alternatives", "top [category]". Open each cited roundup. If you are not on it, send a short, factual pitch with your one-line value, a feature table and a free trial link.
4. **Check JavaScript dependence** If your product pages only render content after scripts run, ChatGPT’s crawler will not read them. Ensure the key text is in the HTML at load or provide server rendering.
5. **Set up simple measurement** List ten buyer questions. Ask them monthly in ChatGPT, Perplexity, Gemini and Copilot. Log names and links. In analytics, build a segment for referrers like chatgpt.com, perplexity.ai and copilot.microsoft.com.

## Allow the answering crawlers, block what you do not want trained

Assistants will not cite you if their search crawlers cannot read your site. Training crawlers are separate. You can permit search while limiting training.

- Answering and search: Googlebot, Bingbot, OAI-SearchBot, Claude-SearchBot, PerplexityBot, Applebot, DuckAssistBot, Amazonbot, cohere-ai, Perplexity-User and Claude-User fetch on demand, and ChatGPT-User fetches specific pages when asked.
- Training controls you may choose to disallow: GPTBot, ClaudeBot, Google-Extended, Applebot-Extended, CCBot, Bytespider, meta-externalagent, meta-externalfetcher.

> Blocking answering crawlers like OAI-SearchBot, Claude-SearchBot, PerplexityBot, Bingbot or Googlebot makes your site invisible to assistants.

```robots.txt
# Example: allow search and answering crawlers, limit training
User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

User-agent: Claude-SearchBot
Allow: /

User-agent: ClaudeBot
Disallow: /

User-agent: Googlebot
Allow: /

User-agent: Google-Extended
Disallow: /

User-agent: Bingbot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Applebot
Allow: /

User-agent: Applebot-Extended
Disallow: /

User-agent: CCBot
Disallow: /

```

Robots.txt is honoured voluntarily. Blocking a training crawler will not remove pages already collected. Keep this file under version control and document why each token is set.

## Build a quotable page that assistants can trust

Assistants prefer pages that answer early, match the question in headings, and show a date and an author. They quote lists and tables. Give them clean material.

- Start with the answer. Example on "/pricing": "Paid starts at £X per month" is not allowed here, so write: "Paid plans include monthly and annual options." Then list features by plan in a table.
- Match the question in an h1. Example: "/alternatives/competitor" with h1 "Best alternatives to Competitor".
- Add a summary list in the first 100 words. Example: three bullet points on who it suits, main features, and migration steps.
- Use consistent names. If your product is "Acme Calendar", do not switch to "Acme Schedules" in half the pages.
- Add a visible date and author. Place the updated date near the title and the author near the end.
- Make features scannable. Numbered steps, comparison tables, and definitions with one line each.
- Link to proof. Case studies, docs, and public status or pricing pages. Assistants prefer verifiable claims.

A fixed page looks like this: a two sentence answer at the top, a short list, a table with labelled columns, then deeper sections with clear h2s like "How billing works" and "Migration".

## Get on the lists and roundups assistants already cite

Being on cited lists is the most reliable route to being named. Assistants often inherit choices from those pages.

1. **Find the lists** In ChatGPT search and Perplexity, ask the buyer questions your market uses. Open the Sources. Make a sheet of the roundups and comparisons they cite.
2. **Decide your angle** Write one line that fits the roundup. Example: "Best for teams that need audit trails" on a "best project trackers" list.
3. **Pitch to be added** Email the editor with a short subject. Three sentences: what you are, why you fit the angle, and one fact they can quote. Offer a free trial link and a feature table. No attachments.
4. **Fix your snippet** On your homepage and "about" page, add a one line description that editors can lift. Keep your name and plan names consistent.
5. **Stack internal links** Link from "/pricing" and "/features" to your new "alternatives" or "best tools" page. Assistants follow these links when they fetch context.

If you sell to local buyers, also claim listings on well known directories in your niche. Assistants read those, then quote the names they see across sources.

## How to measure GEO without an agency

There is no central report of where you were cited. You need a simple loop that is reliable and cheap in founder time.

- Questions list: write ten buyer questions. Example: "best Kanban tools for startups", "Acme Calendar alternatives", "is Acme secure".
- Manual checks: monthly, ask each question in ChatGPT, Perplexity, Gemini and Copilot. Log who is cited and the links.
- Analytics: segment referrers like chatgpt.com, perplexity.ai, copilot.microsoft.com and gemini.google.com. Google AI Overviews and AI Mode arrive as google.com like any search.
- Tracking products: Profound, Peec AI, Otterly, Scrunch, Semrush and Ahrefs AI toolkits, and DataForSEO’s LLM mentions data automate the asking and logging.
- Search Console: clicks from AI Overviews and Google AI Mode are mixed into Search clicks. Watch the Performance report for questions where you also see assistant citations.

A working setup is a sheet of questions, a monthly calendar reminder, a saved analytics segment, and a log of sources gained or lost. Keep it lean so you run it.

## Questions to ask before you pay a GEO agency

- Which answering crawlers are currently blocked on my site and how will you change robots.txt? Show the diff.
- How will you verify that ChatGPT and Perplexity can fetch our key pages without executing JavaScript?
- Which lists and comparison pages do assistants already cite in our niche, and what is your plan to get us added? Name five targets.
- How will you measure citations and share of mentions? Show a sample log with dates, assistants, queries, and source URLs.
- What content will you create in month one? Show a template with the answer-first pattern, headings, date and author.
- What is your plan for our names? Product, plans, and features need consistent terms. How will you enforce that across pages?
- What will you not do? Confirm you cannot buy citations inside assistants and that rankings in Google Search still matter for AI Mode and AI Overviews.
- Who writes the outreach emails and how will you avoid duplicate pitches editors ignore?

If you get vague answers, walk. If they show concrete targets, templates and diffs, you can judge the work on its merits.

## When hiring a GEO agency makes sense

You can ship the basics yourself. Hire when the constraints are scale, time, or coordination across teams and properties.

- You have a large catalogue or many locales. Getting on dozens of cited lists and keeping names consistent needs process and hours you do not have.
- You need outreach at scale with editors who already know the agency. Intros can speed up adds to roundups.
- Your site depends on JavaScript frameworks. You need server rendering or pre-rendering set up without breaking product flows.
- You run several subdomains or apps with separate robots.txt files. You need someone to map and align crawler access.
- You need to produce a run of answer-first pages and comparisons in a quarter. You can brief, but you cannot write them.
- You want monthly monitoring and actioning that your team will not keep up with.

A good agency will start with crawler access, fix a page template, list the target roundups, and show one measurable win in the first cycle.

## How assistants choose and cite sources today

ChatGPT search launched to all users in 2024 and opened without an account in 2025. It uses partners like Bing and its own OAI-SearchBot. It shows inline citations and a Sources panel. It does not execute JavaScript when crawling, so content that only appears after scripts run is not read.

Perplexity searches for every question with PerplexityBot, fetches on demand with Perplexity-User, and prefers pages that answer early with clear headings and recent dates. It shows numbered citations and a sources list.

Google AI Overviews and AI Mode cite pages that Google’s ranking systems surface. There is no separate crawler beyond Googlebot and no opt out beyond snippet controls. Clicks arrive in Search Console like any web Search click, not separated.

Claude search uses Claude-SearchBot and fetches on demand with Claude-User. It shows inline citations. It is reported to build on Brave Search’s index. Microsoft Copilot cites from Bing’s index, with Bingbot its way in. Gemini grounds with Google Search, and Google-Extended controls training and grounding in Gemini apps, not Search ranking.

## Questions

### What is a GEO agency?

A GEO agency helps your site get cited in answers from ChatGPT, Perplexity, Google AI Overviews and AI Mode, Claude, Gemini and Copilot. They do monitoring, write answer-first pages, handle robots.txt and JavaScript issues, and get you onto lists and comparisons assistants already cite.

### Which GEO agency is considered the best?

There is no single winner for every niche. Judge by their plan for your market, the lists they will target, the templates they use, and how they will measure citations. Ask to see diffs to robots.txt and examples of pages that were later cited by assistants.

### Is GEO replacing SEO?

No. GEO overlaps SEO. Assistants still rely on search indexes, and Google AI Mode and AI Overviews cite pages that rank in Search. You still need crawlable pages and strong content. GEO adds work on answer-first structure and getting onto cited roundups.

### How fast can you see GEO results?

You can see a citation in days if you fix robots.txt and land an add on a roundup assistants already cite. Wider gains take cycles. Plan monthly checks across assistants, and expect an uneven curve as pages update and editors add or reorder products.

### What are the top GEO companies?

Lists change often. Use assistants themselves to ask for "best generative engine optimization agencies" and open the sources they cite. Then pick by fit: their experience in your category, the cited lists they can reach, and how they will work with your stack.

### Can you pay to be cited by assistants?

No. There is no paid placement for citations in ChatGPT, Perplexity, Claude, Gemini or Copilot. Agencies can pitch editors of third party lists and improve your pages. They cannot buy mentions inside the assistants.

## Read next

- [Generative engine optimization (GEO): how to be named by AI answers](https://porteur.ai/guides/generative-engine-optimization): How to get your site named in AI answers: sources, crawlers, lists, structure, llms.txt, comparisons, and how to measure by asking assistants.
- [robots.txt for AI crawlers: GPTBot, ClaudeBot, PerplexityBot and what to allow](https://porteur.ai/guides/robots-txt-for-ai-crawlers): Decide which AI crawlers to allow in robots.txt, why it matters, and copy‑paste examples for GPTBot, ClaudeBot, PerplexityBot, Google‑Extended and more.
- [How ChatGPT search picks and cites sources](https://porteur.ai/guides/how-chatgpt-search-cites-sources): See how ChatGPT search finds pages and cites them, what your page needs to be picked, and how to check if your site is named.
- [How Perplexity picks its sources, and how to become one](https://porteur.ai/guides/how-perplexity-picks-sources): How Perplexity finds, indexes and cites pages, what it favours, the crawler tokens to allow, how to see your mentions, and a monthly GEO routine.
- [Google AI Mode: what it is and how a small site gets cited](https://porteur.ai/guides/google-ai-mode): What AI Mode is, how it picks citations, how clicks appear in Search Console, and the page changes that make a small site quotable.
- [How to track your AI visibility without a subscription](https://porteur.ai/guides/ai-visibility-tracking): Track AI visibility without tools: list 20 buyer questions, ask each assistant, log names and links, segment referrers, and know when automation is worth it.
- [Answer engine optimization: the older name for being the answer](https://porteur.ai/guides/answer-engine-optimization): AEO in 2026: what it is, how it overlaps GEO, the page format assistants cite, the crawlers to allow, and how to check if you are named.
- [What AI SEO means, and the three things it can refer to](https://porteur.ai/guides/what-is-ai-seo): AI SEO means three things: using AI to do SEO work, being cited in AI answers, and Google’s AI in Search. Here is what to do on each.

Want a quick read on where you could be cited first, and which crawlers you are blocking, from your URL in about thirty seconds, free? Try the check. Free check: https://porteur.ai/