In a study Semrush published in June 2026, AI answers that cited a brand's page didn't mention the brand in 61.7% of cases. (Semrush) So you can do the hard part, getting your page picked as a source, and still have the reader leave without ever learning your name. Answer engine optimization (AEO) is the work of making your brand the one an AI assistant names, cites or links to when someone asks it a question. The "answer engines" are the AI chat assistants and the AI-generated summaries that now sit at the top of many search results.
In practice, AEO means two things. Your pages have to be easy for these systems to find and quote. And the rest of the web has to describe you the same way your site does.
For B2B SaaS, the questions that matter are usually commercial: "What's the best tool for X?", "What are the alternatives to Y?", "Does Z integrate with our CRM?" The buyer often reads the answer and builds a shortlist before visiting a single vendor site.
Our view, up front: AEO is worth doing for almost every B2B SaaS company, but it's rarely worth a separate vendor or a separate budget. Most of the work is SEO, PR and product marketing you should already be doing, aimed at a new reader. This guide covers how answer engines choose sources (using the most recent studies we could verify), what AEO work involves, how to measure it, and a checklist for your first month.
Contents:
How answer engines choose their sources
The main answer engines today are ChatGPT, Perplexity, Gemini and Google's AI Overviews and AI Mode. No answer engine publishes its full source-selection logic. Google has documented the most, and independent studies fill in some gaps. The evidence from the last year points the same way on most questions.
Most answer engines search first, then write
Google describes two techniques behind AI Overviews and AI Mode. The first is retrieval-augmented generation (RAG), also called grounding: the system uses Google's core ranking systems to retrieve pages from its index, then writes a response from them with links to supporting pages. The second is query fan-out, where the model fires off several related searches at once to gather more material. Google's own example: a query about fixing a weedy lawn might fan out into "best herbicides for lawns" and "remove weeds without chemicals". (Google Search Central)
Fan-out explains a lot. Your page can be cited for a question it never ranked for, because it ranked for one of the sub-questions.
Ranking still helps, but less directly than it did
The March 2026 update covered 863,000 keyword SERPs and 4 million AI Overview URLs. Beyond the 37.9% in the top 10, another 31.2% ranked in positions 11–100, and 31.0% didn't rank in the top 100 at all. (Ahrefs)
Ahrefs notes that it improved its citation parsing between the two studies, so the comparison isn't a clean before-and-after. It points to fan-out as one likely reason for the gap, and notes AI Overviews moved to Gemini 3 in January 2026. Google hasn't confirmed any change to fan-out behavior.
Ranking is still the most reliable way into Google's AI answers, but it's no longer a near-guarantee, and it's not the only door.
Each engine has its own source "diet"
Semrush's 2026 AI Visibility Index looked at 126 million U.S. prompts from January to April 2026 across ChatGPT, Gemini, AI Mode and AI Overviews. ChatGPT cited about 15 sources per response on average. Gemini cited about 3. ChatGPT leaned often on Reddit and Wikipedia. (Semrush)
The overlap between platforms is low. In the full report, citation overlap ran below 50% for every pair of platforms, and Google AI Mode and AI Overviews cited YouTube more than any other source. (Semrush AI Visibility Index report, PDF)
Even one engine isn't consistent with itself. In a Semrush study with Kevin Indig, GPT-5.2 answered an identical set of 100 prompts in both minimal and high reasoning modes. Only 25.6% of cited domains overlapped between the two. The high-reasoning mode leaned more on official documentation and government or academic sources, and less on Reddit. (Semrush)
So when someone tells you "you rank in ChatGPT", ask: which mode, which prompt, which day?
Being cited is not the same as being named
A citation is a link to your page. A mention is your brand name in the answer text. They don't always happen together.
In Semrush's "ghost citations" study (3,981 domain appearances, 115 prompts, 14 countries), Gemini mentioned the brand in 83.7% of appearances yet linked it just 21.4% of the time. ChatGPT did roughly the opposite. (Semrush)
For a SaaS company, the mention is usually the thing you want. A buyer who reads "Tools like Acme and Beta handle this" has your name on a shortlist, whether or not they click.
Off-site signals correlate most with brand visibility
Ahrefs analyzed 75,000 brands to see which factors line up with mentions in ChatGPT, AI Mode and AI Overviews. YouTube mentions had the strongest correlation (~0.737). Branded web mentions came next (0.66–0.71). The number of pages on a site barely mattered (~0.194). ChatGPT showed weaker links to classic authority metrics like Domain Rating (0.266). (Ahrefs)
These are correlations, not causes. Big brands get mentioned everywhere, and they also get mentioned in AI answers. But the pattern is consistent with what you'd expect if models learn about you from what others say, not just from what you publish.
Further reading: AEO vs SEO: what's actually different · AEO agencies for B2B SaaS
What AEO work actually involves
AEO work falls into four buckets, and none of them is exotic.
1. Make sure the engines can reach you
For Google, the bar is plain. Pages must be indexed and allowed to appear in Search with a snippet. AI Overviews and AI Mode add no technical requirements beyond that, according to Google. (Google Search Central)
One further condition comes from Google's newer optimization guide: the site needs to be part of the "Search generative AI" setting in Search Console. (Google Search Central) It's worth a two-minute check.
Other assistants run their own crawlers and search partners. Check your robots.txt and CDN or firewall rules so you aren't blocking the AI crawlers you actually want, and make sure key content doesn't depend on JavaScript that fails to render. Miraa's technical SEO page covers crawl and rendering fixes.
2. Publish pages that answer buyer questions clearly
This is where most of the effort goes. Pick the questions buyers ask an assistant before they talk to sales, and make sure a clear, checkable answer exists on your site.
For SaaS, those questions cluster around a few shapes:
Category and alternatives: what the product is, what else is in the category, and how it differs.
Fit: who it's for, and who it isn't for.
Integrations: what it actually connects to.
Implementation: what happens after the contract is signed, and who does the work.
Pricing and limits: what a buyer can verify.
Google's guidance is to create "non-commodity" content: a unique point of view, first-hand experience, content that goes beyond common knowledge. It explicitly warns that a summary of what's already online "simply restates information already available elsewhere." (Google Search Central)
One rule Miraa works by: if the condition lives three sections away from the claim, a summary will keep the claim and drop the condition. If something holds only on the Enterprise plan, say so in the same sentence. The guide to comparison, alternatives and category pages applies this to "vs" pages.
3. Get described accurately off your site
Because off-site mentions correlate so strongly with AI visibility, AEO extends past your blog: review sites, comparison articles, community threads, YouTube, partner pages, analyst coverage.
The honest version of this work is PR and product marketing. You give people a reason to talk about you, and you correct third-party pages that describe you wrong.
The dishonest version is buying or faking mentions. Google lists "seeking inauthentic mentions" among the things that aren't as helpful as they seem, and notes its AI features depend on the same systems that block spam. (Google Search Central) SEO consultant Glenn Gabe made the same point in September 2026. Reacting to a report that a publisher was testing paid LLM visibility for brand partners, he wrote that if it was selling AI visibility, "that's 'inauthentic mentions'". (Glenn Gabe on X)
4. Keep your brand story consistent
Your product name, category and audience should read the same on your homepage, pricing page, docs, G2 profile and LinkedIn. When the sources disagree, a model has to guess, and it may guess wrong (or describe a competitor more confidently because their story is simpler).
What you can skip (for now)
A lot of AEO advice is unproven. Google's optimization guide has a whole "mythbusting" section for its own AI features. (Google Search Central) For Google Search, it says you don't need:
llms.txt or other AI text files. Google Search ignores them. They won't hurt or help there, though other services may use them.
"Chunking" content into tiny pieces. Google says its systems understand pages that cover several topics.
Rewriting content in a special style for AI.
Special schema for AI. Structured data isn't required, though it still helps with rich results.
Independent data backs up the schema point. Ahrefs tracked 1,885 pages that added JSON-LD between August 2025 and March 2026 against matched control pages. Citations didn't meaningfully rise in AI Mode (+2.4%) or ChatGPT (+2.2%), both within noise, and dipped 4.6% in AI Overviews. (Ahrefs) One caveat: every page in that study was already being cited, so it says nothing about pages AI hasn't found yet.
The bigger warning in Google's guide is about scale. Creating separate pages for every variation of a question, including fan-out queries, "primarily to manipulate rankings or generative AI responses" violates the scaled content abuse policy. (Google Search Central) SEO consultant Lily Ray made the same point on X in December 2025: the spam policies that applied to SEO also apply to the sites Google cites in AI search. (Lily Ray on X) In June 2026 she shared a case where a site's blog got a manual action on May 17, with its citations in AI Overviews and ChatGPT shown alongside. (Lily Ray on X)
A pitch for 500 AI-written "answer pages" carries exactly that risk.
How to measure AEO
Measurement is the weakest part of AEO today. You're sampling a moving target. The setup below accepts that.
Track a fixed set of buyer prompts
Write 20–50 prompts a buyer would actually type, in their words. Run them on a schedule across the engines your buyers use. For each, log three things: were you mentioned, were you cited, and was the description accurate?
Because answers vary by mode and by run (see the 25.6% overlap above), track trends over weeks. Don't react to one screenshot.
Several tools sell this as automated AI visibility tracking, including Ahrefs Brand Radar, Semrush's AI Visibility Toolkit and Similarweb. Google adds a fair warning in its guide that no outside tool can see inside its ranking or AI systems. (Google Search Central)
Use Search Console's Generative AI report
Google introduced a Generative AI performance report in Search Console in June 2026. The report breaks down AI Overviews and AI Mode impressions by page, country, device and date. (Google Search Central Blog) It's the only first-party data you'll get for Google's AI features.
Watch AI referral traffic, and know its blind spots
Similarweb's 2026 Generative AI Landscape data puts AI platform referrals at an average of 770.7 million visits a month worldwide between June 2025 and May 2026, up 117.4% year over year. It's still a low single-digit share of traffic for most sites. (Similarweb)
Two caveats. Similarweb found that after ChatGPT's May 7, 2026 update, around 60% of its referrals landed on homepages, compared with roughly 26–32% earlier. (Similarweb) And Similarweb notes some AI-driven discovery may show up as direct traffic with no clean referrer. (Similarweb)
So segment AI referrers in analytics, but also add "How did you hear about us?" to your demo form. Self-reported attribution catches the buyer who read your name in ChatGPT and typed your URL later. Our B2B SEO pipeline attribution guide covers the full setup.
Tie it to pipeline
Mentions are a leading indicator. Demos, SQLs and pipeline are the result. If you can't connect the two over a quarter or two, revisit which prompts you're tracking.
A practical AEO starting checklist
A first month for a B2B SaaS team looks like this. It's mostly reading and fixing, not publishing.
Week 1: access and baseline
[ ] Confirm key pages are indexed and snippet-eligible in Google Search Console.
[ ] Check the Search generative AI setting in Search Console is set to include your site.
[ ] Review robots.txt and firewall rules for the AI crawlers you want to allow.
[ ] Write 20–50 buyer prompts (category, alternatives, integrations, pricing, "best X for Y").
[ ] Run them across ChatGPT, Perplexity, Gemini and Google, and log mentioned / cited / accurate.
Week 2: fix what's wrong
[ ] List every answer that describes you incorrectly, and find the source it pulled from.
[ ] Fix your own pages first: one clear statement of what you are, who you're for, and what you integrate with.
[ ] Put conditions next to claims (plan limits, regions, prerequisites).
[ ] Flag third-party pages with outdated info (review sites, directories, partner listings) and request corrections.
Week 3: fill the real gaps
[ ] Pick the 5–10 prompts where you're absent but should appear.
[ ] Improve or write one page per gap, with something only you can say: real product detail, real customer outcomes you can name, real data.
[ ] Skip anything you can't support with proof.
Week 4: off-site and measurement
[ ] Identify where competitors who do appear are being mentioned (comparison posts, YouTube, community threads).
[ ] Plan genuine ways to earn a place there: customer reviews, product demos on YouTube, expert contributions.
[ ] Set up the Search Console Generative AI report, AI referrer segments in analytics, and a "How did you hear about us?" field.
[ ] Re-run your prompt set and compare to the week 1 baseline. Expect slow movement.
The short version
AEO isn't a new discipline so much as SEO, PR and product marketing with a new audience: a model that reads everything and repeats what it can justify. Google itself says that, from its side, optimizing for generative AI search "is optimizing for the search experience, and thus still SEO." That's mostly right. The parts that are genuinely new are measurement, which is messier, and off-site reputation, which now matters more.
So hold off on any AI-visibility dashboard until three things are done: key pages are crawlable and quotable, review sites and directories describe you correctly, and a hand-checked baseline has been run of 20 to 30 buyer prompts. To see how the two disciplines split in practice, read the AEO vs SEO comparison. If you'd rather hand it to a team, there are three roundups to compare: B2B SaaS SEO agencies, AEO agencies for B2B SaaS, and AEO agencies built for agentic AI companies. For category-specific playbooks, see SEO for AI and agentic AI companies, sales tech SEO and E-E-A-T and YMYL for fintech. Regulated or comparison-heavy categories have their own lists: fintech SEO agencies and SEO agencies for sales automation software.
Miraa works on AEO for B2B SaaS across ChatGPT, Perplexity, Claude, Gemini and AI Overviews, and reports results as demos, SQLs and pipeline. It doesn't promise citations, because no agency can guarantee a spot in a specific AI answer. Related work sits on its generative engine optimization and LLM SEO pages, and its AEO page describes the approach.