AEO and GEO consulting for companies and agencies
More people now ask an assistant before they search. ChatGPT, Perplexity, Gemini, Claude, Microsoft Copilot, and Google's AI Overviews and AI Mode answer the question directly and cite a handful of sources. I find out what they say about you, where they're getting it, and what would make them say it more accurately and more often.
Who I do this for
I run AI search work for B2B technology companies and support agencies as the search specialist on their client accounts, the way I have on the Meritus Media team. For a company, I own the program: the question set, the fixes, and the reporting. For an agency, I work inside your client accounts under your process, deliver findings your account team can present, and stay out of the client relationship unless you want me in it.
The vocabulary, briefly
The field has too many acronyms for one idea. GEO (generative engine optimization), AEO (answer engine optimization), AIO, LLMO, and AI SEO all describe making a site easy for AI systems to find, understand, trust, and cite. In practice it's good technical SEO, clear entities, and content written so a single paragraph can stand on its own as an answer.
What I test
- A prompt set built from your buyers' questions. "Best X for Y," "Is Z legitimate," "Who does A near B," and comparisons with named competitors, run across the major assistants.
- What each answer says. Whether you appear, how you're described, and whether the facts are current.
- Where the answer came from. The sources each assistant cites, and which of them you control, can influence, or should correct.
- Crawler access. Whether GPTBot, OAI-SearchBot, ClaudeBot, Claude-SearchBot, PerplexityBot, Perplexity-User, Google-Extended, meta-externalagent, and Bingbot can reach your pages, through robots rules, Cloudflare, and your firewall.
- Machine readability. Whether your key content is in the server HTML, since most AI crawlers don't run JavaScript.
What I change
Entities
AI systems reason about things, not keywords: your organization, your people, your products, and your locations. I make sure each is described the same way everywhere, with Organization, Person, Product, and LocalBusiness schema tying them together and sameAs links pointing to the profiles that confirm them.
Answer-shaped content
Pages that open with a direct answer, define their terms, include the numbers people ask about, and keep each section self-contained. FAQs written around real questions. Glossaries. Comparison pages that are fair enough to quote.
Evidence elsewhere
Assistants trust what other sources confirm. I identify the directories, review sites, publications, and community discussions that answers lean on, and where your information there is missing or out of date.
Files for machines
An llms.txt file that summarizes your site and links to clean Markdown versions of key pages, XML sitemaps, and IndexNow, so changes reach Bing, and through it Copilot, quickly.
The first month
The first month produces a baseline and a ranked list of fixes. I start with access: the site or its repository, Google Search Console as an owner, Google Analytics, and Bing Webmaster Tools. I crawl the whole site with Screaming Frog, including a JavaScript rendering pass, and check redirects, broken pages, sitemaps, canonicals, structured data, and what crawlers receive in the first HTML response. Search Console and Bing show which AI-related queries and referrals already exist.
At the same time I run the question set across the assistants and record who gets named, who gets cited, and what each answer says about you. By the end of the month you have the baseline, the gaps between what the assistants say and what is true, and the fixes in order of payoff, written so a developer can act on them.
Measuring it
I measure AI visibility with a fixed set of questions your buyers ask, run every month on the same assistants. Each run records share of voice (the share of answers that name you), your position among the names, the tone of what is said, the sources each answer cites, and every factual claim made about you, checked against the truth. The citation gaps, meaning sources that cite your competitors and leave you out, become the outreach list. Alongside that: AI assistant referrals in GA4 and Bing's AI performance reporting.
My own baseline from October 2026: named in every answer that asked about me by name, and in none of the open hiring questions. That is the gap I am closing, and I report it the same way to clients.
What to expect
Technical access problems can be fixed in days. Changes in how assistants describe you follow their crawl and retrieval cycles, which take weeks to months. Start with my AI search readiness checklist, or see a company that wanted accurate AI answers for job seekers. For the parts of the work you can do yourself, read how to correct what AI says about you, how to earn the mentions answers repeat, and what llms.txt does and doesn't do.
Updated
