What Does a Generative Engine Optimization (GEO) Agency Actually Do?

TL;DR

• Muck Rack found that 84% of the links cited in ChatGPT, Claude and Gemini responses were earned media, across more than 25 million links and 17 industries in its May 2026 edition. A generative engine optimization (GEO) agency that only edits your website is leaving most of the evidence untouched.
• A GEO agency runs a prompt baseline, fixes crawler and indexing problems, restructures pages, makes your brand facts consistent, earns coverage and measures the result monthly, all to get your brand found, cited and described accurately by AI assistants.
• Buyers and AI answers often use GEO and AEO interchangeably, though the terms point at different surfaces: GEO (Generative Engine Optimization) targets LLM-powered platforms like ChatGPT and Perplexity, while AEO (Answer Engine Optimization) targets traditional answer features like Google featured snippets and voice search. AI visibility is the outcome both aim at, and our complete guide to AI visibility compares them.
• The first 90 days of any engagement should buy you a baseline and the start of execution. Early signs, a cleaner site and first placements, can show inside 60 to 90 days; movement in AI answers builds over the months after that, and no agency can promise citations. We would walk away from one that does.

Why does the work have to reach beyond your own website?

An agency improves a brand’s odds of being cited by changing what the engines read, and most of that material sits off your site. Google says its generative AI features “can show what’s being said about products and services across the web”, including blogs, videos and forum discussions.

Third-party coverage gives an engine descriptions it can compare with your own claims. A trade article or an independent review establishes what your product does and who uses it, and a community thread supplies the comparison a company page cannot credibly make about itself. Consistent descriptions across those sources make the evidence clearer when an engine grounds an answer, meaning when it pulls in pages and builds its response from them.

The outreach has to be real. Google warns that seeking inauthentic mentions across the web is not as helpful as it might seem. Useful work looks like pitching informed stories, correcting inaccurate listings and taking part openly in the communities your buyers read; manufactured mentions spread inconsistent claims and hand the engine poor evidence.

That settles the first question to put to any agency: are you changing the sources AI systems read, or only the website? A plan that lives entirely inside your content management system has answered it.

Is a GEO agency different from an AEO or AI SEO agency?

Less than the names suggest. Buyers and vendors often use GEO, AEO and AI SEO for overlapping work, and AI visibility is the outcome all of them aim at. Google’s own guide recognizes AEO and GEO as common terms for work focused on visibility in AI search experiences, and adds that from Google Search’s side, optimizing for generative AI search is still SEO.

The name on the proposal tells you little about whether the agency works beyond your site, so ask that instead.

What are you actually buying?

A GEO engagement typically covers an audit and prompt baseline, site and technical fixes, content restructuring, entity consistency, earned media and measurement. Each should trace its method to Google, Bing or OpenAI documentation, or to disclosed third-party research.

Audit and prompt baseline

A baseline audit records how often your brand appears for a fixed set of buyer questions across ChatGPT, Perplexity, Gemini, Copilot and Google’s AI features. It should separate mentions from citations, record which pages and third-party sources appear, judge whether the descriptions are accurate, and compare your presence with named competitors.

Cross-engine testing matters because the engines cite different sources. How closely each engine’s citations track Google’s top 10 organic results varies widely, from Perplexity at the high end to ChatGPT at the low end (Semrush, July 2025), so strong Google rankings tell you least about what ChatGPT will cite. No first-party report covers every engine; Google’s Generative AI performance report covers only AI Overviews and AI Mode.

A fixed prompt set creates a repeatable sample. We would ask for it in writing, with the wording, the audience context, the engine and the schedule, so the same prompts can be re-run over time. The same prompt can return a different answer on a different day, so any visibility score is a sample with limits, and an agency that presents one as precise has skipped that caveat.

Site and technical fixes

Technical fixes make your existing pages accessible to search and AI crawlers. The agency should inspect robots directives, response codes, canonical tags, internal links and XML sitemaps. Bing’s guidelines recommend IndexNow to tell Bing when URLs are added, updated or removed, and state that NOARCHIVE prevents content from being used in Copilot responses while NOCACHE limits Copilot to the URL, title and snippet.

For ChatGPT, the agency should confirm that OAI-SearchBot can reach the public pages you want surfaced. OpenAI’s publishers FAQ says blocking OAI-SearchBot keeps your content out of ChatGPT’s summaries and snippets, though the ChatGPT Atlas browser may still show just the link and page title. The audit should distinguish OAI-SearchBot from GPTBot, which governs training, because blocking one is not blocking the other.

AI features rely on the same crawl and indexing foundations as search. Google says its search systems do not use llms.txt, need no special schema.org markup, do not require content to be chunked into small pieces, and need no new machine-readable files or AI text files. A standard XML sitemap and accurate structured data for the content already on the page are the deliverables. A proposal built mainly around llms.txt or proprietary AI markup is offering a file in place of that work.

Content restructuring

Content restructuring gives each page one clear purpose and puts its main claims where a reader and a crawler find them first. A restructuring pass moves the direct answer and its evidence near the top, replaces vague references with explicit names, and splits unrelated topics onto their own URLs. Bing recommends focusing each URL on a single topic and placing essential information near the top, because grounded answers depend on content Bing can clearly interpret and verify.

Structure is the cheaper half. Google says that creating content people find unique, compelling and useful will likely influence a site’s presence in generative AI search in the long run more than any other suggestion in its guide. Original research, a firsthand product comparison or an explanation drawn from customer experience gives an engine something it cannot find elsewhere.

Nobody should rewrite your pages solely for AI systems: Google says you do not need to write in a specific way for generative AI search.

Entity consistency

Entity consistency keeps your company name, product names and core descriptions the same across the sources an engine may meet: your website, your structured data, third-party profiles, review sites and earned coverage. Bing advises clear and consistent naming for people, organizations, products and locations, and says clear entity definition improves grounding visibility and citation accuracy.

The agency establishes canonical names and descriptions, then checks external sources for stale positioning, renamed products and factual conflicts. The wording can vary from source to source as long as none contradicts the basic facts about what the company offers, who it serves and how its products relate. Write the canonical description once, put it in your Organization schema, and hand it to everyone who publishes about you.

Earned media and third-party sources

Earned media gives an engine independent sources for checking how a company and its products are described. In the Muck Rack study, 84% of the more than 25 million links cited in ChatGPT, Claude and Gemini responses pointed to earned media, journalism alone made up 27%, and paid and advertorial content accounted for 0.3%.

The service covers pitching journalists with evidence they can use, correcting inaccurate coverage, earning accurate listings on review and reference sites, and helping your subject matter experts take part openly in the communities buyers read. Foundation found Reddit accounted for 20.8% of the top-50 external citation domains across 50 B2B brands, and 30.9% on unbranded discovery queries, the ones where buyers are still building a shortlist.

Traditional PR measures the placement. GEO outreach ties each placement or discussion to a tracked prompt, a product category or a factual description an engine needs to verify, so the pitch starts from a question buyers actually ask. Before we build a media list, we run the category’s buyer questions through ChatGPT, Perplexity and Google’s AI features, record which publications, trade sites and communities the answers cite, and pitch the sources that already appear.

Measurement and reporting

AI visibility measurement combines platform reports, referral traffic and repeated prompt tests, because AI answers do not produce a stable ranking you can watch. Reports should say what each source measures rather than blend everything into one score.

Google’s Generative AI performance report in Search Console counts impressions: how many times links to your site appeared in AI Overviews and AI Mode, by page, country and device. It reports no clicks. Bing’s AI Performance report shows which pages are cited and the grounding queries behind them, from sampled data, and Bing says the view reflects how often pages are cited, not their importance or ranking, and does not represent traffic, clicks or engagement.

Referral data adds the behavior that citation reports cannot show. ChatGPT automatically includes utm_source=chatgpt.com in referral URLs, according to OpenAI’s publishers FAQ, so you can pick out those sessions in GA4 and track whether the visitors complete your key events.

Prompt tracking fills the gaps between engines, and every report should disclose its prompt set, its frequency and its sampling limits. Google states that “no third-party tool has access to our internal ranking or AI systems.” Gauge, the AI visibility measurement platform we use with clients, samples answers across several major engines. Its results show how those sampled answers change over time, not every answer a user receives.

What should the first 90 days buy you?

Month one is diagnosis. The agency audits your technical SEO, content, earned media and community presence, runs the fixed prompt set across the engines, and documents your buyer, their buying triggers, your product language and any conflicting facts about the company before recommending anything. The baseline should be in your hands by the end of that month.

Months two and three start the work: the highest-priority technical fixes, the first restructured pages, the first pitches and open community participation, run from a prioritized plan with an owner on every task. The report at day 90 should separate completed work from sampled change in AI answers.

In our experience, early signs such as a cleaner site, first placements and a growing community presence show up within 60 to 90 days of consistent work, and movement in AI answers builds on those inputs over the months that follow.

What should a GEO audit include?

• A buyer and competitor definition: who you sell to, who you are compared with, and the questions customers put to AI assistants.
• An access list that stops at read-only Search Console. Google’s hiring advice is to grant read access at the audit stage and hold off on write access.
• A technical review covering crawling, indexing, page structure, structured data accuracy, site speed, and access for the relevant search and AI crawlers.
• A content review that names missing topics, unclear answers, unsupported claims, and pages that need firsthand evidence or an expert’s input.
• A PR review that inventories owned content, earned coverage, reference sites, reviews, and the community threads where your buyers talk.
• An AI visibility baseline: the fixed prompt set, run across the engines, recording citations, descriptions, omissions and competitor appearances.
• An entity review comparing your company name, product names, positioning and basic facts across your site and third-party sources.
• A final report that ranks recommendations by expected effect, effort, dependency and owner.
• Realistic estimates of the improvement possible and the work involved, which Google says is what an audit should be about.

What should you ask an agency, and what are the red flags?

Start with the questions Google’s hiring guidance tells you to put to any SEO: previous work and success stories, the results to expect and when, how success is measured, experience in your industry, and whether every site change and the reasoning behind it will be shared with you. Then ask what is specific to this work.

• Which three of our pages would you change first, and why?
• Which recommendations come from official Google, Bing or OpenAI documentation, and which come from your own testing?
• Are you changing the sources AI systems read, or only my website?
• Does the fee cover the audit, the execution, or both, and which technical fixes will you implement rather than recommend?
• Which AI engines and how many prompts will you track, and do the tools cost extra?
• Who, by name, will do the work each week, and can those names go in the contract?
• Who owns the content, data, accounts and reporting history when the engagement ends?

Some answers should stop the conversation.

• Guaranteed citations conflict with Bing’s statement that GEO does not guarantee grounding or citations in AI experiences. An agency cannot control which sources an engine selects.
• A score the agency will not explain leaves you unable to judge what it measures. Google tells buyers to be careful with a company that is secretive or will not clearly explain what it intends to do. Ask for the method in writing. We publish what our own score measures, in four parts: front-end SEO, back-end infrastructure, share of voice and community consensus.
• An llms.txt file as the main deliverable gets you nothing in Google Search. Google’s AI optimization guide says Google Search ignores llms.txt and similar files, and maintaining one for other services will neither harm nor help you there.
• Manufactured mentions can violate platform rules and distort how your company appears. Bing lists link buying, link spamming, private networks and artificial social promotion schemes under abuse, and Google says seeking inauthentic mentions is not as helpful as it might seem.
• AI-generated content published at scale without editorial review creates accuracy, originality and quality problems. Bing warns that large-scale content generated without oversight often lacks usefulness, accuracy and originality, and may be excluded from indexing.

Should you hire an agency or build this in-house?

An agency fits when you need PR, technical SEO, content and measurement covered at once and cannot hire a specialist for each. A single marketing lead who also has to pitch journalists, diagnose crawler problems, revise pages and track answers across several engines is doing four jobs.

Building in-house fits when GEO will be a steady workload and you can staff each function: technical SEO, editorial and PR, and someone who owns measurement. The tooling is Search Console, Bing Webmaster Tools, referral analytics, a media database and a prompt tracker.

An agency still needs you: access to product experts, customer evidence, analytics and executives who will talk to journalists, plus an internal owner who can approve technical changes and sign off messaging. Without that, an outside team cannot produce accurate source material or get work published.

A hybrid keeps product knowledge, approvals and the developer who ships site changes in-house while the agency owns strategy, earned media, community and measurement. Our buyer’s guide to choosing a PR and SEO agency for AI search has a 100-point scorecard you can run on any agency, ours included.

What drives the cost?

An agency’s fee depends on scope, existing technical debt, content volume, the PR effort required and the measurement plan. A site with crawl problems or inconsistent product descriptions needs more specialist hours, producing original content and earning credible coverage takes longer than auditing what exists, and tracking more engines and larger prompt sets adds setup, review and reporting work. Althea Labs prices each engagement after a scoping call, because the right mix depends on your stage, your category and how much your team runs in-house.

The return you can count directly is narrow: sessions from AI assistants that go on to complete your GA4 key events, and leads who say an AI answer sent them. Impressions and sampled answers measure presence. We have not found an independent benchmark for a standard GEO return or payback period, so ask every agency to separate assumptions from guarantees and to report against your baseline.

How does Althea Labs approach generative engine optimization?

Althea Labs is a brand presence and AI visibility agency that builds targeted visibility systems for B2B startups through integrated PR, SEO, and content programs. Althea Labs covers GEO and AEO under a single visibility program, and we call the method PR Plus. We define generative engine optimization as the work of getting your brand found, cited, and described accurately by AI assistants. In every engagement you work directly with Kara Silverman, Althea Labs’ founder.

PR Plus runs the components below as one system, so each makes the others more findable. Two of the four components, Authority and Human Proof, do their work off your site, which is our answer to the first question in this post.

• Authority is traditional PR: earned media placements that build the third-party credibility telling buyers and AI systems your brand is a legitimate option.
• Infrastructure is SEO and schema: crawler access, page titles and descriptions, clear page structure, speed and accurate structured data.
• Human Proof is community: an authentic presence where your customers actually talk, built through genuine participation.
• The Bridge is content: the posts, videos and FAQs that connect your PR authority to your community presence and answer buyers’ questions directly.

The steps run in this order.

1. Audit covers your SEO content and technical SEO, your PR footprint across owned and earned media, and an AI visibility and entity recognition baseline, and ends with a Brand Presence Report with prioritized recommendations.
2. ICP + Messaging defines your buyer and their buying triggers, then builds a messaging framework with a core point of view, proof points and a channel map.
3. Visibility Tracking sets up Gauge and baselines 300 to 500 prompts pulled from real conversations and your own customer interactions; after the baseline, we re-run a tracked set every month.
4. Strategy turns the findings into a 90-day editorial calendar across content, PR and community, with a PR target list and a content pillar map.
5. Execution is the monthly work: writing, PR pitching, social media, paid opportunities, speaking submissions and award nominations, with a monthly visibility report against the baseline. We don’t publish AI content slop, chase algorithms, or use black-box tactics.

FAQs

Do GEO and AEO need different work?

Mostly the same work, aimed at different surfaces. Both need clear information, credible sources and pages an engine can read. Althea Labs covers GEO and AEO under a single visibility program, and our complete guide to AI visibility sets out where the goals and surfaces differ.

How does GEO differ from traditional SEO?

GEO starts where SEO does, with technical health and useful content, then adds what classic SEO leaves out: tracking what AI answers say about you and which sources they cite, and working on how those sources describe your brand. Google says its generative AI features on Google Search are rooted in its core Search ranking and quality systems.

Can an agency guarantee AI citations?

No, and our advice is to be wary of anyone who promises it. Models can return different answers for similar prompts, the platforms control retrieval, selection and generation, and Google says that even for a page meeting every requirement, indexing and serving are not guaranteed.

How is AI visibility measured?

With a fixed set of relevant prompts re-run across engines, plus cited pages, brand mentions, competitor presence and referral traffic. First-party reports are partial evidence: Bing says its AI Performance data is a sample that does not represent traffic, clicks or engagement.

Can a traditional PR firm handle GEO?

A PR firm can carry the earned-media work, and Muck Rack’s May 2026 study found 84% of the links cited in ChatGPT, Claude and Gemini responses were earned media. A complete engagement also needs technical SEO, content work, entity consistency, prompt tracking and analytics, so ask which of those the firm staffs. Our guide to big PR firms and boutiques has four tests to run in the pitch meeting.

Get an audit before you hire anyone

A baseline audit gives you evidence to compare options before you commit to an agency, hire in-house or approve a longer engagement: your current AI visibility, technical barriers, earned media footprint and priority opportunities, so every proposal is held against the same facts.

Althea Labs sells that audit on its own as the Brand Intelligence Audit: five days, delivered as a deck and a one-hour session, covering your SEO footprint, earned media, AI visibility and community presence. Run it with us and then hire whoever you want, including your own team. Start with an AI visibility audit before you choose who executes the plan.