AEO agency: answer engine optimization services for ChatGPT, Gemini and Google AI Overviews

What does an AEO and GEO agency actually do?

Answer engine optimization is the work of making a brand easy for AI assistants to find, understand and name when a buyer asks them a question. Generative engine optimization (GEO) and AI SEO describe the same job under other names. The target is the answer itself, in ChatGPT, Gemini, Perplexity, Microsoft Copilot, Claude, Google AI Overviews and Google AI Mode, not a blue link on a results page. If you are new to the field, our guide to AEO explains the basics, and AEO vs GEO vs SEO sorts out the vocabulary.

Our service has five parts. The audit can be ordered on its own, but the five work best as a cycle: measure, decide, change, measure again.

Audit: where does your brand stand today?

The audit starts from questions, not keywords. We draft the questions a buyer would type into an assistant (for example “which accounting software should a small business choose?”), agree them with you, and ask each one on every engine in scope. In parallel we read your site with the checks the engines document: robots.txt rules for each AI crawler, firewall refusals, noindex and nosnippet directives, text that only appears after JavaScript runs, and structured data that does not match the visible page. The AI visibility audit page details the deliverable. For a first look on your own, the free AEO checker asks three buyer questions live to ChatGPT, Perplexity and Gemini.

Action plan: what should change first?

Every finding becomes an action with an owner, an effort estimate and a status label. “Documented” means an engine's own documentation supports it, and we give the link and the date we read it. “Heuristic” means it comes from our observations or from common practice, and we say so. The GeckoBoost software already proposes part of this plan from the answers it collects: sites to be cited on, pages to write, false claims to correct, searches to rank on. We add judgement: what fits your brand, your budget and your team's time.

On-site work: what do we change on your website?

• Crawler access (documented). We check that Googlebot, Bingbot, OAI-SearchBot, Claude-SearchBot and PerplexityBot can reach your pages, both in robots.txt and at the firewall or CDN, and we fix the rules with your developer.
• Text in the HTML (documented). Important content has to be readable without JavaScript. Bing's guidelines list “hiding critical content behind client-side rendering” among the things to avoid.
• Pages that answer (heuristic). One page per real buyer question, the direct answer in the first lines, a table where it helps, a named author and a visible date. Of the 27 top-three Google pages we analysed on 10 November 2026, 17 name an author, 22 show a date and 15 have a table. An observation, not a documented rule.
• Entity data (documented for the markup, heuristic for the rest). Google says structured data is not required for its AI features, but markup you use must match the visible text. The same name, description and facts on every profile is our heuristic.
• Discovery and freshness (documented). XML sitemaps with accurate lastmod dates and IndexNow submissions, both named in Bing's guidelines.

If you prefer to do this yourself, our guides 10 actions to boost your AEO and how to appear in ChatGPT describe the same steps with their sources.

Off-site work: where else should your brand appear?

Assistants rarely rely on your site alone. In our 10 November 2026 collection, YouTube was cited in 24 of 124 answers and Reddit in 20, ahead of any brand website, while ChatGPT leaned on publisher pages and self-published “best of” lists. Off-site work therefore covers free and verified directories, review platforms, published lists whose editors accept proposals, YouTube and LinkedIn content built around the questions you want to own, and useful answers in communities. These are heuristics drawn from where engines take their sources, not documented ranking factors.

We never buy reviews, votes or hidden links, and we always disclose affiliation. Anything that needs your identity (a company page, a verified profile, a post under your name) stays in your hands: we prepare it, you publish it.

Measurement: did anything move?

Measurement runs in the GeckoBoost software, so you see the same numbers we do. It asks your questions to the engines you choose, once a day on Starter and up to three times a day on Pro and Agency (Claude once a week per question), finds which brands each answer names, flags statements that contradict your fact sheet (Pro and Agency), and shows Google's first page next to the AI Overview. A weekly check reads your website for the access problems listed above and opens or closes the matching actions on its own.

How does an engagement run, week by week?

1. Week 1: audit

   Kick-off call, question list agreed, first collection on every engine in scope, site checks. You give us read access to your analytics and, if you wish, to your site or CMS.

2. Week 2: action plan

   We present the audit and the plan in a review meeting. You decide what we do, what your team does, and what waits.

3. From week 3: monthly cycles

   We work through the plan: access and rendering fixes first, then pages, then off-site work. Each change is logged with its date, so later movement can be read against it.

4. Every month: review

   A report per engine and per question: mention rate, share of voice against the competitors you chose, cited sources, what we changed and what moved. We keep, drop or add actions for the next cycle.

Who does what?

• We do the questions, the collection, the analysis, the plan, the writing, the markup, the off-site preparation and the monthly report.
• You do the approvals for anything published under your brand, the posting on accounts that require your identity, and you name one person who can answer product questions.
• Your developer, or ours, makes the robots.txt, firewall and rendering changes. Without a developer on your side, we make them with access you grant and can revoke at any time.

What do we measure, and in which AI engines?

We measure answers as a person sees them, through each assistant's own interface, because developer APIs can answer differently. Claude is the exception: it is asked through Anthropic's API with web search on. For each answer we keep the text, the brands it names, the sources it cites and, where the engine reports them, the web searches it ran. Every figure in a report can be traced back to the dated answers behind it.

Starter measures up to 4 of these engines, Pro and Agency up to 7. Questions are asked in English, French, Dutch or German, set to the country of the market you sell in. Pro and Agency also check whether the searches the assistants ran show your brand on Bing's first page.

A plan aimed at one engine leaves the others to your competitors. And answers move between collections: on “best AEO agency”, the agency named by four engines on 31 October 2026 was named by two on 10 November. A single measurement is a snapshot. A trend needs the same questions, asked the same way, over several weeks.

Which AEO and GEO recommendations are documented?

Much of what is sold as AEO or GEO is guesswork, as Google itself notes. Every recommendation we make carries one of two labels: documented or heuristic.

What can no AEO or GEO agency promise?

A ranking, a citation or a fixed share of answers. No agency controls how an engine chooses its sources, and the engines write it themselves. Anyone who promises you the top spot in ChatGPT is selling a prediction.

Advertising does not buy answers either. Google states that “paid advertising does not influence the organic responses generated by Google AI” (Google Ads Help, read 31 October 2026), and OpenAI that ads “do not influence the answers ChatGPT gives you” (OpenAI, 16 January 2026).

Who are these AEO and AI SEO services for?

• Brands and SMEs whose buyers now ask an assistant before they search, and who want to know what it says about them.
• Software companies competing on comparison questions (“best tool for…”), where lists and review platforms decide who gets named.
• Marketing teams that already run SEO and need someone to own the AI answer side, with numbers they can show internally.
• Agencies that want to offer AEO or GEO to their clients without building the measurement. We can work under your brand, and the Agency plan of the software produces white-label reports with your logo.

We work in English, French and Dutch, and each question is asked in the language of the market it measures. One local detail matters: in Dutch, “AEO” mostly refers to the customs status of authorised economic operator. Pages written for Dutch-speaking buyers should spell out “answer engine optimization” in full, and ours do.

How much do AEO services cost?

Services are on quote. The work depends on the number of questions, engines, languages and markets, and on how much your own team takes on. A quote follows a 30-minute demo in which we look at your current answers together. The software that carries the measurement has public prices.

The software starts with a 14-day free trial on one project, without a card, with a preview of the results. Plan details are on the pricing page. Our guide to AEO and GEO pricing gathers the ranges other agencies publish, each with its source and date.

What proof can we show today?

We have no client case study yet, and we will not invent one: a client story is published only with the client's written consent. What we can show is our own brand, measured with our own method. On 10 November 2026, GeckoBoost was named in none of 124 AI answers about its own category. The GeckoBoost case study starts from that zero and records, at each measurement, what we published and what moved.

Once a market holds enough answers, the barometer publishes which brands the assistants name in that market, with a margin of error. Who we are, and the company behind GeckoBoost, is on the about page.