How to Build an AI SEO Strategy: 10 Steps to Rank and Get Recommended in AI Search

For most of my career, getting found online meant one thing: a spot on the first page of Google. Then the ground moved in three steps. Google put AI Overviews on top of its US results in May 2024, ChatGPT added live web search that October, and in May 2025 Google opened AI Mode, a chatbot built into search. By that summer, a Pew Research browsing study found that people who saw an AI summary clicked a regular search result in 8% of visits, about half as often as people who didn’t see one.

For a software or tech company, this is very concrete. The IT director or founder shopping for a new tool doesn’t open ten tabs anymore. They describe the problem to ChatGPT, Claude or Perplexity, or read the AI Overview at the top of Google, and get a short list of vendors. If your product isn’t on that list, you’re out of the deal, and you’ll never see it in your analytics. That’s the gap AI search visibility is about.

The plain way I’d put it: there’s now a robot doing your buyer’s first round of research, and it picks the shortlist before any human ever visits your site.

I’ve watched this play out across the dozens of software, SaaS and IT companies I’ve worked with as an SEO and AI search consultant, and at my agency, Nine Peaks Media, where my team runs AI SEO for B2B SaaS and technology companies. The pattern is the same everywhere. A first-page ranking used to be the finish line. Now it’s the entry ticket, and the real prize is being the name the AI repeats when a buyer asks it for a recommendation.

A couple of weeks ago, I was walking the founder of an IoT connectivity company through his AI visibility dashboard during his monthly SEO review. He stopped me and said it felt like his company had spent a lot of time and money getting well positioned in Google, while everybody had moved over to AI search, where his brand barely showed up. He was only half right, as you’ll see in the steps below. But he was describing exactly what his buyers were doing.

Google rankings still matter, and you’ll see how much of AI search still runs on them. They’re just not the whole job anymore, and the playbook keeps moving. Roughly every six months, the sources these tools read and the tactics that move the result change again. So before I get into the AI SEO strategy I build for software companies through 2026 and into 2027, it helps to understand how AI search decides which brands to recommend in the first place.

How AI Search Decides Which Brands to Recommend

AI search tools don’t keep a secret list of the best vendors. They take the buyer’s question, turn it into several regular web searches, read the pages that come back, and write a summary of what those pages agree on. If your brand shows up on enough of those pages, described the right way, you get named.

Google says this openly. Its own guide to generative AI features explains that AI Overviews and AI Mode are rooted in its core ranking systems, and it describes the “query fan-out”: the model splits one question into a set of related searches and pulls results for each. ChatGPT works in a similar way. On a call with a healthcare IT client, I typed one of their tracked prompts into ChatGPT and showed them the search it actually ran. It wasn’t their words. ChatGPT had rewritten the question into a shorter keyword, run it on a search engine, and read the blog posts on the first page before answering.

Two words matter here, and I’ll use them through the whole article:

• A mention is when the AI names your brand in its answer, with or without a link.
• A citation is when the AI lists one of your pages as a source it used.

You want both, and they don’t always come together. A brand can be mentioned because a third-party listicle named it, while its own site is never cited.

The fan-out is also why “rank #1 for your main keyword” isn’t a complete AI SEO strategy anymore. An updated Ahrefs analysis, reported in March 2026, found that only 38% of pages cited in AI Overviews also ranked in the top 10 for the same query, down from 76% in its July 2025 study. Part of that drop comes from Ahrefs detecting citations better, so don’t read it as “rankings stopped mattering.” My reading is simpler: Google is citing pages that rank for the related searches it runs behind the scenes, not only the one you typed.

So what does the AI read? For software and B2B tech, I see the same handful of page types over and over: “best X software” listicles, comparison posts, LinkedIn articles, review sites like G2 and Capterra, YouTube videos, and the vendors’ own product and pricing pages. Your AI search optimization strategy has to cover all of them, which is what the steps below are built to do.

The 10 Steps I Use to Build an AI SEO Strategy

This is the order I run it in with clients. The first three steps are the strategy itself: the problem, the prompts and the baseline. Steps 4 to 10 are the work. Skip the first three and you end up with a pile of tactics and no way to tell whether any of them paid off.

Step 1: Write Down the Business Problem Before You Touch a Tool

Start with one sentence about what’s broken in the business, not in the channel. “Get AI visibility” isn’t a problem. “Demo requests dropped for three months while traffic held steady” is a problem. So is “the leads coming in are the wrong buyers.”

Here’s why I start there. With that IoT founder, traffic and impressions were up and the pages were ranking, yet inbound leads had slowed to a trickle over the summer. The fix for that is different from the fix for a brand nobody has heard of. With a healthcare IT client, the problem was the opposite: one of their best-ranking pages pulled in people searching for a patient login screen, and their team told me straight that they were getting more bad leads than good ones. More visibility would have made that worse.

So before anything else, write down:

• The business outcome: demos, trials, qualified leads or pipeline value, not “mentions.”
• The buyer: job title, company size and industry of the person who signs.
• The current bottleneck: nobody knows you, the wrong people find you, or the right people find a competitor first.

If you already have a solid organic program, this sits on top of it. If you don’t, start with the basics in my B2B SEO strategy guide, because an AI SEO strategy built on a weak site has nothing to stand on.

Step 2: Pick 5 to 10 Buyer Prompts From Real Data

A prompt is the full question a buyer types into an AI tool, like “recommend the best EHR integration vendor for a reference lab.” Your prompt list is the core of the whole AI SEO strategy, because every audit, page and outreach email that follows is built from it. Get it wrong and you’ll optimize for questions nobody asks.

My best source isn’t a tool. It’s the client’s own data. For that healthcare IT client, I built the prompts from Search Console, where I could see real people typing long, question-style searches into Google. Contact forms and sales call notes work the same way, because they show how prospects actually describe their problem. There’s no reliable prompt volume data to lean on instead. ChatGPT doesn’t share what people ask it, so any “prompt volume” number you see in a tool is an estimate.

The rules I follow when I pick them:

• Use a commercial template: “recommend the best [product] for [who I am] in [location],” plus comparison prompts (“[you] vs [competitor]”) and alternatives prompts.
• Skip prompts that won’t return a list of vendors: yes/no questions and pure definitions rarely name brands.
• Keep it small: 5 to 10 prompts per client, grouped by topic. Most prompt-tracking guides suggest 20 to 40. I start smaller because every prompt adds roughly a hundred sources to study and a set of pages to write, and a list nobody can act on is just a dashboard. When a mentoring software client wanted to cover professional and medical associations, I added just two prompts and grouped the medical variations together instead of growing the list.
• Have the client confirm them: my first draft is always a guess until someone who talks to buyers says “yes, that’s how they ask.”

Track both branded prompts (your name in the question) and category prompts (no brand name). Branded prompts tell you what AI already believes about you. Category prompts tell you whether your AI SEO strategy will win new buyers.

Step 3: Record Your Baseline in Every AI Platform

Before you change a single page, write down where you stand today. Without a baseline, three months from now you’ll have activity reports and no proof.

For each prompt, I record seven things: whether the brand is mentioned, whether it’s linked or just named, its position in the answer compared with competitors, how competitors are described, the tone, how consistently it shows up across repeat checks, and every URL the AI used as a source. I track ChatGPT, Perplexity, Gemini and Google AI Overviews, because a buyer might use any of them. My team uses Amadora for the daily tracking, with the date range set to the full 30 days, since a 7-day window bounces around too much to trust.

Then I check by hand. I ask the same question in ChatGPT and in Google myself, because tracking tools are slow and sometimes wrong.

Expect your AI search visibility baseline to be uneven, topic by topic, not just platform by platform. A communications API client of mine had solid AI visibility for its healthcare video topics, where it had built backlinks and rankings for a long time. For its newer softphone topic, it was named in a response or two, on some platforms, some of the time. That gap told me exactly where to spend the next quarter, and it set honest expectations with the client about how fast a brand-new topic moves.

Step 4: Make Sure AI Crawlers Can Read Your Pages

AI tools can only recommend what they can read. This step is technical and boring, and it quietly breaks more AI SEO strategies than anything else on this list.

Check these on your money pages:

• Search crawlers are allowed: Googlebot for AI Overviews and AI Mode, and OAI-SearchBot for ChatGPT. OpenAI’s crawler documentation for site owners says sites that block OAI-SearchBot will not be shown in ChatGPT search answers.
• The page returns a real 200 status and is indexed: if Google never indexed the page, no AI tool built on Google will find it.
• Important text is in the HTML: content hidden in accordions, tabs, calculators or loaded by JavaScript after the page opens is often missed. I move anything I want quoted out of accordions.
• Your brand name is written the same way everywhere: the full company name, not a nickname, so the AI doesn’t split your mentions between two spellings.

Two myths to clear up while you’re in there. Blocking Google-Extended in robots.txt doesn’t take you out of AI Overviews; the controls Google lists for its AI features are nosnippet, data-nosnippet, max-snippet and noindex, so check none of those sit on pages you want quoted. And you can skip llms.txt files. Google’s guide says plainly that Google Search ignores them, and I haven’t seen them move anything for any of my clients.

Step 5: Reverse-Engineer the Sources AI Reads for Each Prompt

This is the step most AI SEO strategies skip, and it’s the one I’d keep if I could only do one. For every tracked prompt, pull the full list of pages the AI read, sort them by type, and you have your to-do list.

The lists get long. For the healthcare IT client, the tracked prompts added up to about 1,000 URLs that the AI tools read, around 100 per prompt. My team sorts every URL into buckets: listicles, comparison posts, the client’s own pages, competitors’ pages, LinkedIn, review sites, YouTube, forums. Whatever dominates the list decides the content mix. For most software categories it’s listicles. For a few, it’s the vendors’ own service pages, and then chasing listicles is wasted effort.

Most source-gap guides stop at “see which domains get cited.” The useful part is the decision it forces: which type of page you need to build or get onto, prompt by prompt, so your AI SEO strategy follows the sources instead of a template.

Then look at the fan-out for Google AI Overviews. I send the client’s real questions to Google’s Gemini API, which returns the actual searches it ran to build the answer. For an online booking software client, one keyword broke into 47 separate searches. Each of those is a search your pages can rank for, and a reason the AI might pick a competitor instead of you.

This is also where your assumptions die. That IoT founder couldn’t understand why several companies he’d never heard of were ahead of him in AI answers. When I opened the sources, the answer was right there: the ones I checked had published their own “best IoT connectivity” listicles on their own sites, got them ranking organically, and the AI kept reading them.

Step 6: Rewrite Your Money Pages So AI Can Quote Them

Your product and service pages are the pages you control completely, so fix them first, before you spend a dollar of your generative engine optimization strategy on anyone else’s site. The goal is a page where any single section can be lifted out and still make sense: the answer first, then the detail.

I run every money page through two passes. The first is for ChatGPT, Perplexity and the other chat tools: what’s missing compared with the pages they cite, and what would make them trust the brand. The second is for Google AI Overviews, built from the fan-out searches in Step 5. For a video conferencing client, I ran both passes on the same page and merged them into one ready-to-paste list of copy changes, new FAQs and new blog topics, so the client reviewed one document instead of two.

What usually goes on the page:

• A plain “who it’s for” section: industries, company sizes and use cases, written in your buyer’s words.
• “Best for” and “not ideal for” lines: these help you win “best X for Y” prompts. Clients resist the second one, so I often put it inside an FAQ.
• FAQs built from real fan-out searches: short, direct answers, the way your CEO would explain it to a buyer.
• How your pricing works: buyers ask AI tools what a product costs, and clear pricing on your own page lets them shortlist you without a sales call.

Test before you roll it out. With the booking software client, I updated one internal page first, for a single business type, to see how it moved before touching the homepage and main category pages. If you’re a SaaS company working through this, my approach to SEO for SaaS companies covers how those money pages fit together.

Step 7: Publish New Content for the Gaps, at a Pace Google Trusts

Every tracked prompt points to topics you haven’t covered yet, and filling those gaps is how you rank in AI search for the long, specific questions buyers actually ask. Paste the prompt into ChatGPT, read the sources, and you’ll find five to ten blog ideas in each one. Some have zero search volume. That’s fine, because the AI still runs those searches.

Three rules I hold to:

• Match the format the AI reads: if the sources are listicles, write a listicle. If they’re how-to guides, write a guide.
• Be honest in your own listicles: put your product second or third, describe competitors fairly, and include who each option is not right for. A list where you’re first and the only link goes to you reads like an ad to Google and to the AI.
• Publish at a steady pace: two to three posts a week is my working ceiling for most client sites. I’d rather wait than trip a quality filter (more on that in the section on what didn’t work).

Put a real author on every post. For the healthcare IT client, I added an author bio with a link to a verifiable profile, because Google and the AI tools both look at who wrote the page.

Be patient with the results. On one cleantech client’s state-by-state pages, my team rebuilt the page structure with state-specific regulations. Texas picked up more impressions after the update. Florida didn’t move at all. Same work, different outcome, which is why you track every page separately.

Step 8: Get Your Brand Onto the Pages AI Already Cites

Your own site is only part of the picture, and this is where an AI search optimization strategy really differs from a classic SEO plan. Most of what the AI reads about your category lives on other people’s sites, and the fastest way into the answer is to get added to the pages it’s already reading.

Take the listicles and roundups from your Step 5 source list and reach out to each publisher to ask for inclusion. With a mentoring software client, I send the list to the client first for approval, because some of those pages cover features or audiences that aren’t a good fit. When a listicle ranks well, one placement does three jobs: an AI citation, a backlink for regular SEO, and referral traffic from people who actually read the list.

Expect a low reply rate and plan for several rounds of follow-ups. Some publishers ask for a link exchange or a fee; I treat a paid insertion on a list that already ranks as a fair trade, and I skip lists that get no organic traffic at all, because the AI rarely reads them. Also cover:

• Review sites: G2, Capterra and the category directories that show up in your sources. A profile with zero reviews doesn’t help.
• Comparison and alternatives pages: if the AI keeps reading “[competitor] alternatives” posts, you need to be on them.
• YouTube: it shows up in AI source lists more than most B2B marketers expect, especially for how-to topics.

Step 9: Publish on LinkedIn and Medium, Where AI Already Looks

This is the step that surprised me most over the last year, and it’s the cheapest AI visibility win on the list. LinkedIn articles show up in AI source lists constantly for B2B topics, and they’re far easier to publish than a guest post. A Semrush study of LinkedIn citations across ChatGPT Search, Google AI Mode and Perplexity found LinkedIn was the second most-cited domain, appearing in 11% of AI responses on average.

I see it in my own client data. For the communications API client, LinkedIn articles published just over a month earlier were already showing up in the sources, and one of them had been used 58 times across different prompts. For the healthcare IT client, one AI Overview named them first, pulling from their blog post plus a couple of their LinkedIn pages: three places, three mentions, one answer. And it’s not only for AI. A construction software client got referral visits from LinkedIn articles, and one of those visitors booked a demo.

How I run the LinkedIn part of an AI SEO strategy:

• Publish long articles, not quick posts: the same topic list from Step 7, written as full LinkedIn articles and on Medium.
• Pick the right profile: company pages and real people with an active history. For listicles that name competitors, I publish from a team member’s profile so it isn’t tied to the client’s founder.
• Get them indexed: Medium posts can sit unindexed for a long time before Google finds them. I submit the URLs and share them from LinkedIn to speed that up.
• Space them out: two or three a week across platforms, with a different topic on each, so it reads like a real publishing schedule and not a content dump.

Step 10: Keep Ranking in Google, Because AI Still Reads Page One

Here’s the part that founder was missing. Every blog post and backlink my team had built for his site over the previous four months was also feeding AI search. The listicles AI reads are mostly the ones that rank. The pages ChatGPT reads after its hidden search are the ones on page one. His AI visibility had moved from zero to around 5%, then 10%, because of that organic work, not in spite of it.

So the organic program keeps running underneath your AI SEO strategy:

• Backlinks from sites with real traffic: I check that the linking site gets organic visitors before I care about anything else. Here’s how I approach link building if you want the details.
• Rankings for the fan-out searches: not just your head term, but the related searches from Step 5.
• Internal links: from every new post back to the money page it supports.

The Ahrefs data in the first section points the same way: Google is citing pages that rank for the related searches, not only the main one. Cover the whole topic and you show up in more of them.

How I Use AI Tools Inside the Work Itself

A good share of the articles ranking for “AI SEO strategy” are really about using AI tools to do SEO faster. That matters too, so here’s how I actually use them, with the limits I’ve hit.

• Audits and briefs: Claude does most of the heavy analysis for my AI visibility reviews, because it can run longer and handle more complex jobs than the other tools I’ve tried. It takes the prompt data, sources and fan-out searches and turns them into page recommendations and content briefs.
• Outreach research: my team built a Claude skill that takes the list of listicles AI cites for a client, finds the editor, researches the site and drafts and sends the inclusion email. A person still reads the replies and handles the negotiation.
• Writing: AI writes first drafts from an outline built on real call notes and data, then a person edits it and checks it against the client’s facts.
• Checking AI’s own advice: auto-generated action plans from tracking tools are roughly 60% right in my experience. I’ve seen one recommend schema that was already on the page and a “hub page” the homepage already was. Every recommendation gets checked against the live site before it reaches a client.

One more practical limit: ChatGPT’s deep research quietly stops partway through big jobs. Ask it to check every item on a long list and it checks part of it and calls the job done. Run big jobs in batches.

How to Measure AI Search Visibility

Measure mentions and citations separately, every week, and tie them back to the business problem from Step 1. AI visibility goes up and down like rankings do, so look at the trend over months, not a single check.

What I track:

• Visibility per prompt: how often you’re mentioned, your position in the answer, and which of your URLs are cited.
• Google’s own data: Search Console now has a generative AI performance report showing impressions and pages from AI Overviews and AI Mode. Google started rolling it out in June 2026 and made it available to all sites by the end of August.
• Bing and Copilot: Bing’s AI Performance dashboard in Webmaster Tools shows how often your pages are cited in Copilot and Bing’s AI answers, and the searches behind them. It doesn’t show position, so read it as a trend.
• Referral traffic: visits from chatgpt.com, perplexity.ai and similar sources in your analytics.
• Leads: add “where did you hear about us” to your forms, because a lot of AI research ends with the buyer Googling your name, which shows up as branded search.

Watch quality, not just volume. On one client site I reviewed recently, visitors from ChatGPT had about a 60% engagement rate, higher than the site’s organic traffic. There just weren’t enough of them yet to produce leads, which is useful to know before you judge the channel.

Review Your AI SEO Strategy Every Quarter

Things change too fast for a yearly plan. Every quarter, I sit down with the prompt data and ask four questions:

1. Which prompts moved, and which pages or placements explain it?
2. Which sources did the AI start or stop reading?
3. Which tactics should I drop, because the platform changed under them?
4. Is the business problem from Step 1 still the right one?

Then cut prompts that never return a vendor list, add new ones from recent sales calls, and shift effort toward whatever the source lists now favor. Over the past year, my answer to question three changed twice, which leads to the part most guides leave out.

What Didn’t Work and What I Stopped Doing

Not everything I tried to rank in AI search worked, and some of it backfired. Here’s what I stopped.

Self-promotional listicles at volume. Through 2025, the go-to tactic was publishing “best X” lists with your own brand first, at scale, with titles like “best tools for 2026, updated.” My own agency’s site was one of the sites doing it, and Google’s January 2026 update hit exactly that pattern: rankings, impressions and traffic dropped together. Now the client goes second or third, the list is honest, and the links don’t all point back to one brand.

Publishing too fast. Ten posts a week, week after week, looks like a red flag to Google, and when the next core update came, sites doing that got dropped. Two to three a week is where I hold most clients now.

Reddit seeding. For a while, Reddit threads were everywhere in AI answers, and getting brands mentioned in them worked. By mid-2026 it stopped being worth it. Comments were removed without notice, new accounts hit karma walls and cooldowns, and a friend who works at an AI visibility tool company told me his accounts were banned and the communities he’d built were deleted. I stopped active Reddit work and put that effort into LinkedIn, Medium and listicle outreach.

Trusting the tools’ action plans. See the 60% figure above. Following an auto-generated plan line by line can create duplicate pages that compete with each other.

Expecting ads to fix it. I’ve started testing ChatGPT ads for one client. In the first week, the visitors were real, and 20 to 30% of them stayed on the site for a minute or more, but there were no leads yet. It’s too early to judge, and paid placement doesn’t replace being the brand the AI recommends on its own.

Frequently Asked Questions

What Is an AI SEO Strategy?

An AI SEO strategy is a plan for getting your brand found, cited and recommended by AI search tools like ChatGPT, Google AI Overviews, AI Mode, Perplexity, Claude and Gemini. It starts with a business problem, a short list of buyer prompts and a baseline, then covers your own pages, third-party sources and measurement.

Is AI SEO Different From Regular SEO?

It’s built on regular SEO, and Google itself says optimizing for its AI features is still SEO. The difference is what you add on top: tracking prompts instead of only keywords, studying which sources the AI reads, and getting your brand onto those third-party pages. If your organic foundation is weak, fix that first with solid B2B SEO work, because AI search reads the same index.

How Long Does It Take to Rank in AI Search?

With a solid AI SEO strategy in place, for long, specific prompts, I usually see the first mentions in about three to four weeks after the right pages and placements go live. Broad category prompts take months, because they depend on authority and rankings you build over time. New topics for an established brand still start slow.

Do I Need to Rank on Google to Show Up in ChatGPT?

Mostly, yes. ChatGPT runs web searches behind the scenes and reads what comes back, and the listicles it reads are usually ones that rank. Ranking isn’t a guarantee of being mentioned, but without it you’re relying on luck.

Which AI Platforms Should a B2B Software Company Track?

For AI search visibility, track ChatGPT and Google AI Overviews for every client, then add Perplexity, Gemini and AI Mode as your buyers use them. Claude sends almost no referral traffic for most of my clients today, but the work that helps you in ChatGPT and Perplexity usually helps there too.

Can AI Tools Do AI SEO for You?

They can do a lot of the research and first drafts. They can’t decide which prompts matter to your sales team, check whether a recommendation fits your live site, or build the relationships that get you onto a publisher’s list. Every AI-generated plan I use gets checked by a person.

Key Takeaways

• Start with the business problem: demos, lead quality or awareness, written in one sentence before you pick a tool.
• Build prompts from real data: Search Console, contact forms and sales calls beat any prompt volume estimate.
• Record a baseline: mentions, citations, position and sources for every prompt, on every platform.
• Study the sources: the pages AI reads for each prompt are your content and outreach plan.
• Fix your own pages first: answer first, readable in HTML, with real FAQs and pricing.
• Get onto the pages AI cites: listicles, review sites, LinkedIn and Medium articles.
• Keep ranking in Google: AI search still reads page one and the related searches behind it.
• Measure and review every quarter: drop tactics that stopped working, the way I dropped mass listicles and Reddit seeding.

Conclusion

An AI SEO strategy isn’t a new set of tricks. It’s the same discipline of figuring out what your buyers search, building pages that answer it and earning a place on the sites that matter, applied to a buyer who now asks a robot for the shortlist first. The companies that win are the ones that measure where they stand, fix what the AI can’t read, show up on the pages it already trusts, and keep adjusting as the platforms change.

If you’d rather have someone build this with you, I work as an AI SEO consultant with a full team behind me, and you can book a free strategy session to look at your AI SEO strategy and AI search visibility together. I mostly help B2B SaaS companies, construction and fintech software vendors, healthcare IT companies and IoT connectivity providers get recommended in ChatGPT, Google AI Overviews and Perplexity.