How AI answers choose which companies to cite
There is less mystery here than the market suggests. Google states there are no additional requirements or special optimizations to appear in AI Overviews or AI Mode: a page must be indexed and eligible to show with a snippet. Google also describes a query fan-out technique, issuing related searches across subtopics, which is why a strong page on a narrow sub-question can be cited even when you do not rank first for the head term.
ChatGPT search and Perplexity work from their own crawlers. OpenAI says sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers, and that GPTBot, the training crawler, is separate. Perplexity recommends allowing PerplexityBot for its search results. A surprising number of B2B sites block one or more of these by accident through CDN bot rules.
- Allow OAI-SearchBot and PerplexityBot if you want to appear in those engines; decide on GPTBot separately
- Make sure key content is in the server-rendered HTML, not only injected by JavaScript
- Use IndexNow to notify Bing and other participating engines when pages change
What makes a B2B page citable
AI engines quote statements that are specific, self-contained and attributable. A page that opens with a two-sentence direct answer, names the use case, states constraints and shows who wrote it gives the engine something safe to lift. A page that opens with market context and ends with "contact us to learn more" does not.
Google's guidance on helpful content says trust is the most important element of E-E-A-T. For B2B that means named authors with real roles, customer evidence, dated updates and honest limits. Comparison and alternatives pages matter disproportionately, because buyers ask engines to compare vendors and engines look for sources that do it fairly.
- Lead with a 40 to 60 word answer, then support it with detail
- Publish comparison, alternatives, integration and use-case pages, not just thought leadership
- Keep company facts identical across your site, LinkedIn, review sites and directories
Measuring AI visibility without fooling yourself
AI answers vary by phrasing, location, login state and the day you ask. A single screenshot proves nothing. We track a fixed set of buyer prompts, typically 50 to 150 for a B2B category, run them on a schedule across engines and record citations and recommendations over time. The trend line is the signal, not any single run.
Google traffic from AI features is not split out separately: Google says it is included in overall search traffic in Search Console under the Web search type. So we pair prompt tracking with referral analytics for AI engines and, most importantly, with CRM attribution. Aavenir, a CLM platform on ServiceNow, now sees 90-95% of its results from inbound, and qualified meetings moved from single digits to tens per month.
Set expectations by engine. Answers built from pages fetched at query time can reflect a fixed page or a newly opened crawler rule quickly. Answers that lean on what a model already knows change slowly, because they depend on how often and where your brand has been mentioned across the web over months. That is why off-site citations and consistent entity facts sit alongside on-site work in every program we run.
- Freeze the prompt set for a full quarter so month-over-month comparisons are fair
- Record which third-party domains each engine cites for each prompt, then target those sources
- Tag AI referrers as their own channel in analytics and in CRM source fields
- Report citation share next to pipeline, never screenshots on their own
Common AEO and SEO mistakes
The most expensive mistake is chasing traffic that cannot buy. HiddenBrains grew from 50K to 600K monthly users in 11 months, which shows what a disciplined content engine can do, but for most B2B companies the goal is the few hundred buyers who matter, not the biggest audience. Map every page to a buying question first.
The second mistake is relying on tactics that have stopped working. Google no longer shows FAQ rich results in Search, so FAQ markup alone will not win visibility. Structured data still helps machines understand a page, but the answer itself has to be on the page and worth quoting. A page that says something specific, current and genuinely useful to a buyer will outlast every formatting trick the market invents next year.
- Do not publish AI-generated pages at volume without expert review
- Do not let product, pricing-model or integration pages go stale; engines cite outdated facts too
- Do not measure success on rankings alone when the summary above them decides the click
AEO, GEO and AI search terms, defined
AI search introduced a set of new terms, and vendors use them loosely. These definitions are the ones we use in audits and reports, so the numbers mean the same thing to marketing, sales and leadership.
- Answer engine optimization (AEO): structuring content and brand signals so engines can extract and present a direct answer from your page.
- Generative engine optimization (GEO): improving the odds that large language model responses cite or recommend your company.
- AI Overviews and AI Mode: Google's generated summaries and conversational search, which draw on indexed pages eligible for snippets.
- Citation: a link or named source an AI answer shows as supporting evidence.
- Citation share: the share of tracked prompts where your domain is cited, compared with named competitors.
- Recommendation rate: the share of tracked prompts where the engine names your company as an option.
- Prompt set: a fixed list of buyer questions tracked across engines over time.
- Query fan-out: Google's method of running related sub-queries to assemble an AI answer, which rewards pages covering sub-questions.
- AI crawlers: bots such as OAI-SearchBot and PerplexityBot that fetch pages for AI search; blocking them removes you from those answers.
- Entity consistency: identical company facts, such as name, category, locations and products, across your site, LinkedIn and directories.
- AI referral sessions: website visits arriving from AI assistants, tracked as their own channel.
Common mistakes when buying AEO and SEO services
Most failed AI search programs fail for reasons that are visible in the first month. Watch for these patterns, whether you run the work in-house or with an agency.
For a sense of what content-led search can do, IQLECT built $6.4M of qualified pipeline from organic search and technical content, and Aavenir drew 90 to 95% of its results inbound.
- Changing the tracked prompts every month, which makes progress impossible to read.
- Blocking AI crawlers by accident through a firewall, CDN rule or old robots.txt entry.
- Publishing dozens of thin pages on keyword lists instead of a few pages that answer evaluation questions in depth.
- Leaving comparison, pricing-model and integration questions unanswered, so engines cite competitors or review sites.
- Letting company descriptions differ across the website, LinkedIn and directories.
- Reporting rankings and sessions without connecting AI referrals and organic visits to opportunities in the CRM.
- Expecting pipeline in weeks and cutting the program before content has been indexed and trusted.
- Ignoring the third-party sources engines already cite for your category, such as review sites, analyst pages, podcasts and industry media, where a single accurate mention can matter more than another blog post.
- Treating AI visibility as a separate project owned by nobody in sales or product marketing.
