.png)
AI search is restructuring how buyers discover brands. The companies that show up in ChatGPT, Perplexity, and Google's AI Overviews aren't necessarily the ones with the strongest websites. They're the ones most widely mentioned across the sources these models trust.
This guide breaks down exactly how to earn those mentions.
Estimated time to complete this process manually for 20 target prompts: 40–60 hours.
Want the tracking spreadsheet? We built a free template with every tab, column, and scoring framework referenced in this guide. [Download the LLM-SEO Tracker →]
Why AI Search Changes Everything About SEO
Phase 1: Map Your Target Prompts Phase
Phase 2: Audit What LLMs Already Say
Phase 3: Score and Prioritize Your Sources
Phase 4: Build Your Publisher Hit List
Phase 5: Outreach at Scale
Phase 6: Negotiate, Execute, and Deliver
Phase 7: Verify and Optimize Over Time
The Honest Truth About This Process
A Faster Path
SEO teams have spent decades optimizing what they can control: their own website. Stronger content, better technical foundations, more backlinks, more demand captured.
LLMs don't care about any of that… Not directly, anyway.
The research backs this.
*For 22 more compelling AEO stats, check out our blog post here.
The point is, your website alone cannot secure your place in AI search. Your domain authority won't save you. Your content volume won't either.
What matters is whether your brand shows up in the articles and guides that these models already trust.
Here's how to do it.
You're not optimizing for keywords anymore. You're optimizing for the questions people ask LLMs.
Start by listing the prompts where you believe: "If someone asks this, my brand should be part of the answer."
Open a spreadsheet and brainstorm the main "money" questions in your category:
For each question, write 3–5 variants.
People phrase things differently. "Best project management tools" and "which project management tool should I use for a remote team" may produce very different LLM answers. Small wording changes can completely reshape which sources the model pulls from, and honestly, we don't fully understand why yet.
Add three columns to your sheet:
Question,
Intent (comparison, how-to, problem→solution),
and Priority (1 = critical for revenue, 3 = nice-to-have).
You'll keep revisiting this list. New phrasing will surface from sales calls, customer tickets, social threads, and competitors you didn't know existed. There are emerging tools that help generate likely prompts from your URL, but expect this to be an ongoing, evolving list, not a one-time exercise.
Now you need to see what LLMs are actually saying today and which sources they're leaning on.
Each model displays citations differently. You'll end up with a big, messy list.
Trust the process; that’s the raw material for everything that follows.
For each priority question from Phase 1, run it through ChatGPT, Perplexity, and optionally Claude or Gemini.
For every response:
Don't stop at LLM citations.
Expand your list via Google. LLMs draw on the same ecosystem that Google surfaces to humans, so for each question, search it in an incognito window and collect URLs from the top 10–20 organic results.
Focus on:
Ignore homepages, pure product pages, and obvious AI-generated spam.
Add every URL to your spreadsheet with a Discovery Source column (LLM citation vs. Google SERP) and the SERP position if applicable.
We’ve created a plug-and-play AI prompt to help you brainstorm all the possible prompts your customers may be searching for to find you. Copy it here.
Or if you’d prefer to use an AI visibility tracker tool, here are our top picks.
You now have a sprawling sheet of URLs. Time to turn it into a ranked target list.
De-duplicate by URL.
Then add these columns and fill them in for each page:
Open each URL and skim enough to answer: Is this actually about the question? Does it mention my brand or competitors?
Fill in the columns.
You'll be opening dozens of pages (eventually hundreds), reading just enough to label them. It's easy to lose track or double-work tabs you've already reviewed. Discipline with your spreadsheet is the only thing standing between you and a mental breakdown.
Then assign a simple priority score:
Filter to score ≥ 2.
That's your hunting ground.
You know which pages matter, but now you need to figure out who controls them and what to ask for. This is when the real work begins.
Finding contacts: For each high-priority URL, look for the author byline, author bio, or the site's "About" / "Contact" page. Search "[Author Name] + [Site Name] + LinkedIn" or use tools like Hunter, Apollo, or RocketReach for emails.
If you can't find the author, look for the Head of SEO, Content Lead, or Marketing Manager at that company on LinkedIn.
Add columns: Contact Name, Contact Role, Contact Email, Confidence (High/Med/Low).
For each URL, decide:
Tag each URL with an Ask Type: listicle_add, listicle_upgrade, correction, add_recommendation, or other.
Now the unglamorous part: emailing all these people in a way that doesn't feel like spam. (It will sometimes feel like spam anyway. The point is to not be incredibly annoying.)
The core challenge is personalization at scale.
Over-automate and you sound like every other link-building email that publishers have learned to ignore.
Hand-edit every message and you'll burn through hours on outreach alone… with not much to show for it.
Here’s how to find the middle ground:
Create 2–3 core outreach templates covering your main ask types (listicle add, correction/update, general inclusion request).
Each template should:
Load your CSV into a cold outreach tool (Smartlead, Apollo, Mailshake, etc.) with personalization tokens: {{first_name}}, {{site_name}}, {{article_title}}, {{suggested_snippet}}.
Set a light cadence: initial email plus 1–2 follow-ups over 7–14 days.
Expect reply rates around 1–2%.
The reality is that you're sending a lot of messages to earn a few responses. If that feels discouraging, we get it. But it's the honest math of cold publisher outreach.
And they are the gatekeepers to LLM visibility.
You’re going to get some people who will straight up ignore you, others who will say no and, if you’re lucky, a few hesitant yes’s.
The key here is keeping track of all of these replies with different publishers. Each will have different expectations, timelines, and terms.
Handling replies.
Classify each response quickly:
Negotiating terms: Know your boundaries going in.
What's your budget for paid placements?
Are you willing to swap mentions (you add them to a resource page on your site, they add you to their article)?
Any brand compliance constraints?
For reverse mentions, identify pages on your own site where you could reasonably mention them (partner pages, resource roundups) and propose a clear trade.
Be aware that reverse deals create two workstreams: their site and yours. Managing both simultaneously across dozens of publishers is its own project.
Delivering content: When someone agrees, make yourself the easiest person they deal with that day.
Re-read their article. Match their tone. Write 1–3 sentences that fit their structure, not a generic blurb you paste everywhere. Suggest exactly where it belongs: "This would fit after [Competitor X] in your list" or "This could work as a bullet under your 'Tools for [use case]' section."
Fulfilling your obligations: If you agreed to reverse mentions, add them to your site promptly.
File tickets with your web team if needed. Confirm the live pages. Send a quick "all set on our side" note.
Nothing kills a publisher relationship faster than slow follow-through on your end of the deal.
A placement doesn't count until it's live, indexable, and (ideally) getting picked up by LLMs.
Verify every mention. Maintain a tracking tab: Publisher URL, Expected Anchor Text, Expected Target URL, Status (pending / live / broken). When a publisher confirms a placement, open the page and check: correct link, reasonable context, actually visible.
Monitor LLM impact. Every few weeks, re-run your core prompts through ChatGPT and Perplexity. Look for your brand appearing more frequently, and for new citations that include pages where you've earned mentions.
A note on timing: there's typically a lag between a mention going live and LLMs incorporating it.
Training schedules vary, retrieval systems update on their own timelines, and frankly, we don't know exactly how long the delay is for any given model. It could be days; it could be weeks.
The point is: don't panic if a placement doesn't show up in AI answers immediately.
Build your playbook. At least monthly, review what's working:
Use these patterns to refine where you focus, how you pitch, and how you structure offers going forward.
The compounding effect here is real, and it's the reason this work is worth the grind.
Each earned mention makes the next one easier to justify ("we're already featured in X, Y, and Z"), and LLM visibility snowballs as your brand shows up across more of the sources these models trust.
The more painful the manual review becomes, the more it means you're building something.
If you've read this far, you understand two things:
This approach works. Getting your brand mentioned in the sources LLMs cite is the most direct lever for AI search visibility.
It's also brutally manual. Each of those seven phases involves real, repetitive, human work. Spreadsheet wrangling, copy-pasting across LLM interfaces, cold outreach with single-digit reply rates, negotiation across dozens of threads, content drafting in other people's voices. Internal coordination, ongoing verification, etc.
A realistic estimate: 10+ hours of work to earn a single confirmed mention. Multiply that by the number of prompts you want to win.
So the question becomes: Does your team have the capacity to run it at the scale needed to make a real difference?
We built Noble because we went through this exact process and realized the strategy was right but the execution couldn't scale.
Noble compresses the seven manual phases into a handful of steps:
You keep the strategy and control. Noble handles the spreadsheets, the cold emails, the follow-ups, the payment logistics, and the verification.
With a 10% publisher conversion rate (5–10x higher than the 1–2% typical of manual outreach) and up to a 20% increase in LLM visibility, the work reliably turns into live, compounding brand mentions.
→ Get started at thatsnoble.com | → Watch how it works
Want to skip the manual work? Book a demo call with us →
GET DISCOVERED ON