Answer engine optimization for growing companies

Why doesn’t AI recommend my company?

There’s rarely one reason, and a single answer won’t tell you which. When ChatGPT, Gemini, Perplexity or Google’s AI Mode uses web search for a question like “who are the best ERP consultants for mid-sized manufacturers,” it can turn that question into several related searches, read what comes back, and write an answer from those pages. So a common reason is that your company isn’t discussed on the pages those searches turn up, next to the problems you solve. Other reasons include how the question was worded, the buyer’s location, whether the assistant searched at all, a site that’s hard for crawlers to read, and ordinary variation from one answer to the next.

The work has three parts. Find out which sources show up in the answers for your category. Get your company discussed accurately on the ones that matter, in connection with the problems you actually solve. And make your own site clear enough that when an assistant or a buyer lands there, it backs up what they heard.

How AI answers use web sources

A model can answer from patterns it learned during training, which stop at a cutoff date. Many assistants can also search the web when you ask, and use what they retrieve to build the answer, often with links to the sources. ChatGPT may decide to search on its own when a question needs current or specific information. Google says AI Overviews and AI Mode can run several related searches behind one question, which it calls query fan-out. Each product works differently, and the details change often, so treat this as a general picture.

Questions about vendors, like “who should I hire for this,” tend to be the kind that need current, specific information. When the assistant searches, the pages it retrieves shape the answer a great deal. Training still affects how a model understands your brand in general.

A few practical consequences:

• Live retrieval doesn’t mean instant. An assistant searching at question time is reading what’s in a search index, so an edit to your site shows up only after the page has been crawled and indexed again.
• Access matters. If your site blocks the crawlers an assistant uses for search, your own pages can’t be among its sources. Content that only appears after scripts run may also be missed by some of these crawlers.
• Your website is one source among many. For a lot of vendor questions, the cited pages belong to publications, directories, communities and comparison articles.

A prompt is a different thing from a keyword

With an assistant, people talk the way they’d talk to a colleague. A buyer might ask, “We’re a mid-sized manufacturer and our current inspection equipment keeps failing audits, who should we look at?” That reads a lot more like a prompt than a search term.

Prompts are longer, they carry context about the buyer, and they vary a lot. That’s why I don’t think there’s a single playbook. In practice, that means testing a set of questions built for your buyers instead of borrowing a generic list.

How I start: ask the questions, then read the citations

A good starting point is simple enough that you can do it this afternoon. Go to a few AI tools and ask the kinds of questions your target buyer would ask, without mentioning your brand. See what comes back. More importantly, dig underneath and look at the citations, where the answer is pulling from.

Do that across a range of prompts for your ideal customer and patterns start to show. In most categories, a handful of sources come up again and again in the citations. Your job is to figure out how to get your brand into those places, talking about the problems you actually solve.

I did exactly this before a follow-up call with a healthcare services company. I ran sample prompts for their space and traced where the citations came from. LinkedIn articles were pretty heavily used, and Reddit was very heavily used in their space. That gave us something concrete to talk about on the call.

Two cautions about the method. A citation map shows what turned up in the answers you sampled. It doesn’t reveal how a platform weighs sources internally, and it doesn’t prove that appearing on a given site will get you named. Frequent citations to a site tell you where the conversation is happening, which is where planning starts.

To make the sampling worth something:

• Record the product, the date, and the location and account settings you tested with. Location and saved memory can change what ChatGPT searches for.
• Separate the outcomes. Being mentioned, being recommended, having your page cited as a source, and being described inaccurately are four different results. A company can be recommended on the strength of someone else’s page, and a page can be cited without the company being recommended.

The sources differ by category. In some it’s trade publications and industry associations. In others it’s review sites, directories, a couple of big comparison articles, or YouTube. You only find out by looking.

What AI answers are doing to search traffic

On a discovery call, I pulled up a prospect’s search visibility report. His leads were down by about a third year over year. The strange part was that his articles were still pinned to the first position in Google. They sat below the AI Overview, the map pack and everything else on the page, so the top ranking was earning very little traffic.

Before blaming the AI Overview for a drop in leads, you’d want to separate rankings, clicks, conversions, overall demand and any tracking changes. Still, I told him I didn’t think it was just him. A lot of businesses have seen something similar with informational, top-of-funnel content that an AI Overview can answer right on the results page. People often get what they need there and don’t click through, even when the source is cited.

I also asked him whether those articles had been driving business forward in the first place. Sometimes traffic is down and the business is fine. That’s worth knowing before you spend money chasing traffic back.

When answer engine optimization is worth paying for

Not every company needs ongoing work here. The questions I use to decide:

• Do your buyers use AI assistants to research vendors? Your sales team and your recent customers can usually tell you.
• What do the assistants say about your category today? If you’re missing, described wrong, or consistently replaced by the same competitors, there’s something to work on. If you already show up well, light monitoring may be enough.
• Are your SEO basics in place? If your important pages aren’t indexed or are hard to crawl, fix that first. It helps everywhere.
• Do the economics work? A few large deals a year can justify more effort than a high volume of small ones.

The answer after a review can be substantial work, a check a few times a year, fixing existing SEO, or nothing for now. Each of those is a useful answer.

What still matters about your own website

Your site still matters, though its job has shifted.

People aren’t making $100,000 purchasing decisions just because ChatGPT told them so. They still talk to people and look at your website. A buyer might first hear your name in an AI answer, a directory, a social platform or a referral, and land on your site to check you out. Your site can also put you on the shortlist in the first place. Either way, it has to help them judge fit: clear about who you serve, specific about what you’ve done, and easy for both a person and a machine to read.

How do you show up in Google’s AI Overviews?

Google’s AI Overviews and AI Mode are part of Google Search and draw on its index, so classic SEO matters more here than anywhere else in this field. Google says there’s no special markup or extra optimization required for its AI features. A page needs to be indexed and eligible to appear with a snippet in regular search. The pages it cites tend to be pages it already finds relevant for related searches, though not always the ones at the very top.

What helps, in my experience: pages that answer a specific question plainly and early, the way a person would, with the detail that proves you know the subject. Clear headings that match how people ask. Proof that’s specific, like real examples and real numbers, rather than claims. And a site Google can crawl without trouble.

Is SEO dead now that AI answers the question?

No. AI answers are changing how people search, including inside Google itself, and the clicks that used to come from informational content are thinning out. Existing SEO work still matters, and much of it feeds AI answers too.

The way I’d frame it: there are motions that help both, and there are things to do very specifically for AI visibility. Good content, clear site structure and a strong reputation help everywhere. The AI-specific part is examining the answers and citations across assistants, then letting what you find steer your content, SEO and PR. If you already have a solid SEO program, answer engine optimization should extend it.

The best way to frame any organic activity, in my view, is to only ever do things that would help your brand as well. If a tactic would embarrass you in front of a buyer, it isn’t worth whatever visibility it buys.

AEO, GEO, AI SEO: does the name matter?

Not much. Answer engine optimization, generative engine optimization and AI SEO are all names for the same goal: showing up inside the answers that AI assistants give. Agencies use whichever term their buyers search for. What matters is whether the work starts from your actual prompts and citations.

What an answer engine optimization review includes

When I scope this work, it usually covers six things:

1. A question set. A set of questions your buyers would plausibly ask, written with your sales team, covering problems, comparisons and “who should I hire” questions. It’s a useful starting sample, not a full picture of what buyers ask.
2. A baseline. Those questions run across the assistants your buyers use, more than once, recording whether you’re mentioned, recommended or cited, how you’re described and who appears instead.
3. A citation map. The sources that showed up in those answers, ranked by how often.
4. A technical check. Whether the crawlers assistants use for search can reach and read your site.
5. Recommended changes to your own site. Clearer statements of who you serve and what you’ve done, on the pages that matter most, prioritized and written so whoever does the work can make them. Making the changes is separate unless we agree otherwise.
6. A plan for third-party sources. Which publications, directories, communities or partners are worth showing up in, what you’d need to say there to earn a mention, and what isn’t worth the effort.

Then we set a date to rerun the questions and look at what moved. Whoever does the work, the plan and the review date stay the same.

How to measure it without fooling yourself

Visibility is measurable. Run the same core questions across the main assistants on a schedule and track how often you’re mentioned, recommended and cited, how you’re described, which competitors appear, and which sources get cited. Tracking tools such as Peec AI automate the repeat runs, and a rough version by hand works to start. When something moves, treat it as a finding to look into. It doesn’t automatically mean the changes caused it.

Attribution is the hard part. You can track visits that arrive from AI assistants in your analytics and see what those visitors do. Those numbers undercount, because a lot of AI-driven discovery happens when a buyer reads an answer, remembers a name, and searches for you later. So I’m careful with claims. In the TDI work, the case study records the company appearing in AI answers for its category and its content being cited as a source. That’s an observation. Nobody can say visibility on its own produced a specific number of leads, and I’d be wary of anyone who claims that kind of precision.

Supporting signals worth watching, each with plenty of other possible causes: branded search over time, direct traffic, what new leads say on the first call when you ask how they found you, and whether sales conversations start further along.

How to choose someone for answer engine optimization

Some questions worth asking any AEO agency or consultant:

• “What prompts would you track for us, and how did you pick them?” If they haven’t asked about your buyers, they’re guessing.
• “Which sources get cited in our category right now?” A good answer comes from looking, not from a general list.
• “What won’t you do?” Anything that would put your domain or your reputation at risk should be off the table.
• “How will you report on it, and what won’t you be able to prove?” You want someone who’s clear about the limits.

This work tends to connect to the website and to how clearly the brand says what you do. Clear, consistent information gives both buyers and assistants better evidence to work from.

Based in Houston, working nationwide

I’m based in Houston and advise companies across the U.S. on how they show up in AI answers. Location matters less here than in traditional local search, because ChatGPT and AI Overviews usually answer the question, not the zip code. Where it does matter (a Houston buyer asking for a local provider), I’ll show you how the answers differ and what feeds them.