A generative engine optimization agency is a firm that works to get your brand named, quoted and linked inside AI answers from ChatGPT, Perplexity, Google AI Overviews and similar engines, rather than only ranking your pages in classic search results. The research that named the practice, the GEO paper from Princeton and partners presented at KDD 2024, found that adding citations, quotations and statistics to a page lifted its visibility in generative engine answers by up to 40% in their benchmark, so a good agency should be able to show you which of those changes it makes and how it measures the effect.
Key Takeaways
- A GEO agency should measure brand mentions and citations inside AI answers on a fixed prompt set, not just keyword rankings and traffic.
- Most of the work is the same as good technical SEO and content work, plus crawler access for AI bots, quotable page structure and third party mentions the engines already trust.
- Paid placement in ChatGPT is now a separate lever from earned citations, so decide early whether you want one partner for both or two.
If you want both the paid and the earned side covered, see how Adsome handles ChatGPT Ads and AI search content.
What is generative engine optimization?
Generative engine optimization (GEO) is the practice of shaping your content and your wider web presence so that AI answer engines pick your brand and your pages when they compose an answer. A generative engine is a search product that writes an answer from several sources instead of returning a list of ten links. An answer engine citation is a link or named source that the engine shows next to, or inside, that written answer.
GEO overlaps heavily with SEO. Google says plainly in its own documentation that there are no extra requirements to appear in AI Overviews or AI Mode beyond being indexed and eligible to show with a snippet. What changes is the unit of success. In classic SEO you win a position. In GEO you win a sentence: your brand named in the answer, your page cited as the source, your product on the shortlist. If you want the background on how engines make that choice, we covered it in how AI search engines choose which brands to recommend.
What does a GEO agency actually do?
A competent GEO engagement usually covers six workstreams. Ask any agency you speak to how it handles each one, and in what order.
- Prompt set and baseline. Write 50 to 200 real buyer prompts for your category ("best running shoes for flat feet under 150 euros", "is brand X worth it") and record what each engine answers today, who is named and which URLs are cited.
- Crawler access. Check that AI crawlers can reach your site. OpenAI separates OAI-SearchBot (used to surface sites in ChatGPT search), GPTBot (training) and ChatGPT-User (fetches made on a user's behalf). Perplexity documents PerplexityBot and Perplexity-User. Blocking the search bot in robots.txt or at the CDN quietly removes you from those answers.
- Quotable page structure. Rewrite key pages so each answers one question in its first two sentences, defines terms in one line, and puts specs, prices and comparisons in tables that a model can lift without reinterpreting.
- Evidence on the page. Add sources, named standards, real figures and short expert quotes. These are the methods that performed best in the GEO paper, while keyword stuffing performed poorly.
- Off-site mentions. Get your brand into the places engines already cite for your category: review sites, comparison articles, forums and trade press. This is closer to digital PR than to link building.
- Measurement and iteration. Re-run the same prompt set on a fixed schedule, track share of answers and citations per engine, and tie changes to specific pages and mentions.
A good starting point for step 1 is our guide on how to audit what AI says about your brand category.
How we compared the options
We compared the four ways an ecommerce brand can get this work done, using the criteria below.
- Measurement. Whether the option produces a repeatable answer share and citation report across engines.
- Content capacity. Whether it can actually rewrite and publish pages, not just recommend.
- Off-site reach. Whether it can earn mentions on third party sites the engines trust.
- Paid side. Whether it can also run ChatGPT Ads, which is now a separate buying channel.
- Control. How much of the knowledge stays inside your company if you stop.
We did not compare prices. Agency fees vary widely by scope and market and are rarely published, so ask for a scoped quote against the six workstreams above.
GEO agency vs SEO agency vs in-house vs tools
| Option | Measurement across engines | Writes and publishes pages | Earns off-site mentions | Runs ChatGPT Ads | Knowledge stays with you |
|---|---|---|---|---|---|
| Specialist GEO agency | Yes, core deliverable | Usually | Usually, via PR | Sometimes | Partly, ask for the prompt set and raw data |
| SEO agency adding GEO | Varies, often Google only | Yes | Sometimes | Rarely | Partly |
| In-house team | Only with a tool or scripts | Yes | Hard without contacts | Yes, if trained | Fully |
| Monitoring tool only | Yes | No | No | No | Fully, but nobody acts on it |
Tools tell you where you stand but change nothing. A specialist agency earns its fee only if it does the off-site work and hands you the data.
Hard facts each engine publishes
These are the documented rules and research results worth checking an agency against. If a pitch contradicts them, ask why.
| Source | What it says | What to ask the agency |
|---|---|---|
| GEO paper, KDD 2024 | Citations, quotations and statistics lifted visibility by up to 40% on its benchmark | Which of these do you add, and to which pages first? |
| OpenAI crawler docs | OAI-SearchBot controls inclusion in ChatGPT search, GPTBot is for training | Have you checked our robots.txt and CDN rules for both? |
| Google AI features docs | No special markup or extra requirements for AI Overviews and AI Mode | Why would we need anything beyond solid SEO for Google? |
| Perplexity crawler docs | PerplexityBot indexes, Perplexity-User fetches on request | Are we reachable by both? |
What we see in our own ChatGPT Ads account
A GEO agency works on earned visibility. ChatGPT now also has a paid route, and the two should be planned together. These are our own observations from the ChatGPT Ads account Adsome has run since September 2026. The platform is new and changes often, so treat them as what we saw in September 2026 and check OpenAI's own documentation for the current rules.
- The ad format is a chat card: a title of 3 to 50 characters, a body of up to 100 characters, and a square image.
- Targeting in an ad group is by context hints, free-text themes that describe the conversations an ad should appear next to. There are no keywords to bid on.
- A campaign budget is either a daily limit or a lifetime limit.
- Ads are shown only to people on the Free and Go plans. Plus, Pro, Business and Enterprise users do not see ads.
- A custom audience needs 25,000 matched users before it can be used for inclusion. There is no minimum for exclusion.
- Sweden and every EU market could be targeted from a European ad account.
- Conversions can be sent server to server through a Conversions API, for events such as a booked appointment, a new lead or a completed registration.
One lesson from our own account: a lifetime budget can be spent in full on the first day, after which the campaign still shows as active but cannot deliver. Check spend, not status.
Why this matters for choosing a GEO agency: context hints are written like the buyer prompts in your GEO baseline. The same prompt research feeds both the content plan and the ad groups, so an agency that does only one side leaves half the research unused. Adsome is a Stockholm studio that runs ChatGPT Ads and produces the content that helps brands get cited in AI answers, which is why we plan the two from one prompt set.
Questions to ask before you sign
- Which engines do you measure, and how often do you re-run the prompt set?
- Do we own the prompt set, the raw answers and the citation data if we leave?
- Which pages will you change in the first month, and why those?
- How do you earn third party mentions, and on which sites for our category?
- How do you separate the effect of your work from model updates the engines ship on their own?
- Do you run ChatGPT Ads, and if not, who will?
If the answers lean on proprietary scores you cannot reproduce, be careful. AI answers vary between runs, so ask how many times each prompt is sampled before a change is reported.
Which option fits your situation
- Small DTC brand with one marketer. Start in-house with a fixed prompt set and the on-page fixes in our guide to building brand authority so AI tools cite you. Bring in help for off-site mentions once the basics are done.
- Brand already paying an SEO agency. Ask them to add multi-engine measurement and crawler checks. Our AEO vs SEO comparison explains where the two overlap.
- Brand in a crowded category where competitors own the answers. A specialist GEO agency with real PR reach is worth it, because the gap is usually off-site.
- Brand that also wants paid reach in ChatGPT. Pick one partner for both, or make sure the two share the prompt research. For a wider view of agency types, see best AI ad agencies and alternatives by brand fit.
Common Mistakes
- Buying a dashboard and calling it GEO. A visibility score with no pages changed and no mentions earned is a report, not a programme.
- Blocking AI crawlers by accident. CDN bot protection often blocks OAI-SearchBot and PerplexityBot by default. Check the logs, not only robots.txt.
- Measuring one sample per prompt. Answers shift between runs. A single screenshot proves nothing either way.
- Rewriting every page at once. Change a defined set, re-measure, then expand, or you will never know what worked.
- Treating paid and earned as rivals. Ads and citations answer the same prompts. Plan them together.
Want your brand in front of buyers while they ask ChatGPT? Adsome runs ChatGPT Ads and builds the content that gets brands cited in AI search.
