Hello Reader
You've invested in SEO, built a content engine, improved your website and worked on converting more visitors into customers.
Now you need more people coming through the door.
Except there's a problem.
Organic traffic isn't what it used to be.
In January 2025, Search Engine Land reported a dramatic decline in HubSpot's organic search traffic. Traffic fell from an all time high of 24.1 million in 2022 to 8.6 million in December of 2024.
Now that buyers are doing their research inside AI tools, they’re not visiting our websites.
Which means getting found in Google is no longer enough.
You also need to understand how your brand shows up in ChatGPT when a buyer asks it to help them evaluate their options.
And because everyone is trying to figure this out there’s a growing industry selling marketers advice on how to make that happen.
- Add FAQ sections.
- Rewrite your articles.
- Publish on Reddit.
- Get more backlinks.
- Fix your schema.
- Buy another analytics platform.
Whether they call it Answer Engine Optimization (AEO), Generative Engine Optimization (GEO), or simply SEO for AI, everyone has an opinion.
But what’s actually working?
I've been digging into the research and talking with marketers who are testing these approaches to find out.
1. Identify when and how AI recommends you
When you ask ChatGPT a question, it may generate several search queries to find information before writing its answer. This process is called query fan-out, and those searches are known as grounding queries.
The queries AI searches for aren't necessarily the words in your original prompt, and if your website doesn't appear in those search results, it has little chance of being cited in the final response.
Charlie Treadwell, CMO at Elisity, has built a system to track this.
He looks at three things: whether Elisity appears in AI's search results, whether its content gets cited, and whether the brand is mentioned in the final answer.
Try this:
1. Identify the questions buyers are asking.
Start with sales conversations, RFPs, customer interviews and common product comparisons. Choose five to ten prompts relevant to your category.
2. Check whether AI is searching the web (or just relying on it’s training data).
Some responses use live search, while others rely on the model's existing knowledge. If the model isn't searching and your brand is missing, changes to your website won't immediately influence that response.
Focus on questions where you can improve the information AI retrieves.
3. Find out what AI is searching for.
The search queries AI generates may differ from the buyer's original question. Tools such as Profound can help reveal these queries, so you can understand what information the model is looking for.
4. Check whether your content appears in the results.
If it doesn't, investigate your SEO visibility and whether your content uses the language buyers and search engines recognize. If it does appear, check whether the page directly answers the buyer's question.
5. Check citations and recommendations separately.
- Did AI cite one of your URLs?
- Did it mention your brand?
- Did it recommend your product?
Track all three to understand where you're losing visibility.
What we know: AI search still relies on many of the same foundations as traditional SEO. Your content needs to appear in relevant search results, answer the buyer's question and give AI a reason to cite it.
That said, a search result, a citation and a brand recommendation aren't the same thing. Measure each separately to see where you need to improve.
2. Make sure AI understands what you sell
AEO/GEO consultant Andrea Lechner-Becker has seen a familiar B2B marketing problem showing up in AI search.
Companies invent new categories and proprietary terminology to differentiate their products, but buyers don't necessarily use those terms when researching solutions.
"Everybody read that they should be creating a category,” she says. “So now they’re using a bunch of words that no one else uses to search for their shit and of course, they’re not showing up."
If customers describe a problem differently from the way you describe your product, AI might find more relevant information about your competitors.
Founder of CMO Huddles, a B2B CMO community, Drew Neisser says companies need to be particularly careful when describing themselves across their websites, profiles, and other channels.
"If you have different descriptions of your company in different places,” he says. “The LLMs will all come back with different answers."
He recommends creating a short, consistent description of the business and using it consistently across all your channels.
Try this:
Ask ChatGPT, Claude and Gemini:
- What does [Your Company] do?
- Who is [Your Company] best suited for?
- What problems does [Your Company] solve?
- Who are [Your Company]'s main competitors?
- When would you recommend [Your Company] over those competitors?
Then review your homepage, product pages, LinkedIn company profile and leading directory listings.
Are you consistently describing what you sell, who it's for and the problems it solves? Are you using the words customers use?
What we know: Clear positioning and consistent product information give AI systems better information to work with. They don't guarantee recommendations, but inaccurate descriptions and unfamiliar terminology are obvious problems worth fixing.
3. Give AI evidence beyond your website
Your website is only one source of information about your brand.
AI search systems can also use customer reviews, industry publications, analyst reports, product comparisons, directories and community discussions.
That means what other people say about your company can influence how AI represents it.
"One of the things that we meticulously focus on for content is external links that we trust," says Andrea.
Her approach includes using credible sources to support the claims made in content.
In a GEO study, researchers from Princeton tested ways to improve content visibility in generative search. Adding relevant statistics, quotations and credible citations improved visibility up to 132% in some cases.
Try this:
Return to your AI recommendation audit and examine the sources supporting your competitors.
Look for:
- Independent product reviews and comparison articles.
- Industry publications and analyst reports.
- Relevant customer discussions and review platforms.
- Research, statistics and case studies cited by other websites.
Identify which sources repeatedly appear and whether your brand has comparable coverage.
Then review the claims you're making on your website to see if you’re supporting them with credible evidence.
A claim that your product improves productivity is less useful than a customer case study showing which process changed, how it changed and what the customer achieved.
What we know: Research supports including relevant data and credible sources in content. Independent coverage can also provide additional evidence about a brand, but there's still limited evidence about which specific third-party activities consistently increase recommendations across commercial AI systems.
4. Create content that answers buying questions
One mistake marketers make is treating GEO as another reason to produce more content.
We've spent years building articles around keywords, but repeating that process with chatbot prompts creates an enormous amount of content without necessarily answering the questions buyers are asking.
Instead Charlie identifies which questions prospective customers might ask AI about Elisity's products.
"I don't use the words that most people use, and they use phrases that I think are very bizarre."
Charlie has built an AI agent using 20 previous RFPs, including the original buyer questions and Elisity's responses.
When he asked Claude to identify questions missing from his existing search research, it consulted the RFP information and surfaced additional product requirements.
Those requirements gave him new ideas for the questions buyers might ask AI.
Anita Moorthy, Co-founder and CMO of Rocksalt, surveyed B2B marketers to understand which AEO activities are delivering pipeline results and while the survey is still in the field there are patterns emerging.
"Teams publishing or updating 20+ pieces a month were the most likely to tie AEO to measurable pipeline, while only 1 in 11 low-volume publishers could,” she says. “Teams with no single owner couldn't measure any pipeline impact at all."
Her recommendation is to continuously identify the questions buyers ask in sales calls, support tickets, Reddit and LinkedIn. Then answer them with original, factual content on your website and across relevant third-party platforms.
You don't necessarily need to publish 20 pieces a month, but you do need someone accountable for AEO, a repeatable process for finding buyer questions, and a consistent publishing cadence.
Try this:
Choose one high-priority product or solution page.
Ask AI to review the page against five questions your customers routinely ask before buying.
Look for missing information about:
- Who the solution is designed for.
- Which problems it solves.
- How it compares with alternatives.
- Relevant integrations, requirements and limitations.
- Pricing considerations and evidence of results.
Add the information buyers need, particularly details that competitors explain more clearly.
Correct product names, industry terms and acronyms. Turn valuable webinar sections into standalone articles that answer specific buyer questions.
Put your experts to work.
You don't need a massive following or a C-suite title to get noticed by AI. Meltwater and LinkedIn's research found that 75% of LinkedIn citations came from individual members, compared with just 25% from company pages.
Identify two or three people inside your company who know your products and customers inside out. Encourage them to publish practical answers on LinkedIn, using firsthand examples, data and comparisons. Focus on information buyers can't easily find elsewhere.
Then track whether their content starts appearing as a source in your AI recommendation audit.
What we know: Research supports clear, evidence-rich content. It's not clear that mass-producing chatbot-targeted articles delivers lasting commercial results. Focus on the questions that influence buying decisions.
5. Measure recommendations (not just rankings)
Steven Loewy, VP Marketing at Marq says the biggest lesson he learned with AI search is starting with measurement.
“Before we had measurement in place, I was reacting to the topic du jour stakeholders were reading on X and LinkedIn,” he says. “Getting visibility into AI search has helped us compete against much larger competitors because we can make agile, very specific updates to our content strategy.”
Traditional SEO gave marketers familiar metrics:
- keyword rankings
- impressions
- organic sessions
- conversions
AI recommendations are harder to measure.
A brand might appear in ChatGPT but not Claude. It might be cited as a source without being recommended. It might appear for one buyer question but be absent from another.
Here’s what the marketers I talked to recommend you track:
- Recommendation frequency: How often does your brand appear across the buyer questions you're testing?
- Competitive visibility: Which competitors appear more frequently, and for which questions?
- Source visibility: Which websites and articles are cited, and does your content appear among them?
- Business results: Are AI-referred visits, qualified inquiries and opportunities increasing?
Drew suggests repeating your assessment every three months because websites, content and external coverage keep changing.
What we know: Recommendation frequency is a useful discoverability metric, but it isn't proof of business impact. Use consistent questions, repeat your tests and connect visibility changes to commercial results wherever possible.
Which AI search platforms should you monitor?
Not all AI search tools work the same way, and your brand may appear in one but be missing from another.
Here are the major platforms worth including in your visibility audit.
- ChatGPT — Buyers use it to research products, compare vendors and build shortlists. ChatGPT can search the web and cite sources, but it can also answer from its existing knowledge.
- Google AI Overviews and AI Mode — AI Overviews appear within traditional Google results, while AI Mode supports more complex research and follow-up questions. Both draw on Google's search infrastructure, making existing SEO performance particularly relevant.
- Perplexity — Built around AI-powered search, with source citations central to the experience. Useful for understanding which publishers, reviews and comparison pages AI references when researching your category.
- Gemini — Google's standalone AI assistant. Buyers can use it to compare products and research purchasing decisions. Test Gemini separately from Google Search because the experiences and recommendations differ.
- Claude — Particularly relevant for detailed B2B research, document analysis and vendor comparisons. Claude can search the web, but not every response involves a live search.
- Microsoft Copilot — Worth monitoring for B2B brands, particularly those selling to Microsoft-heavy organizations. Copilot can use Bing to retrieve web information and support its answers.