Back to blog
AI & GTM Engineering

How to get recommended by AI: my weekly tracker and 4 actions

Xavier Caffrey
Xavier CaffreyNovember 5, 2026 · 9 min read
How to get recommended by AI: a weekly tracker across ChatGPT, Claude, Gemini and Grok, and four actions that raise how often AI names you.

To get recommended by AI, find out what ChatGPT, Claude, Gemini and Grok already say when your buyers ask them for a company like yours, then work on the sources they cite: write the pages they’re missing, get featured on the sites they quote, fix wrong facts about you and join the threads they pull from. I track this every week, and on Grok the share of answers that name OneAway went from 7.1% on October 24 to 42.9% on November 3.

Ask Grok for the best cold email agency for B2B SaaS and OneAway now comes up at number two. We’re an outbound agency. Cold email and LinkedIn are what we do. But I want to understand every way B2B companies get found, and AI answers are now one of them.

So I built an AEO tracker with Claude Code. AEO is answer engine optimization: getting AI assistants to name you in their answers. This guide is the written version of the video above. It covers what the tracker measures, which numbers matter and the four actions that moved the number.

Quick answer

Write the questions your buyers ask. Put them to ChatGPT, Claude, Gemini and Grok every week, and record who gets named and which pages get cited. Then work through four kinds of action: write a page, get featured, fix a fact and join the conversation. After you publish or edit a page, request indexing in Google Search Console so it gets picked up sooner.


Track four AI engines every week

You can’t improve what you don’t measure, and you can’t measure it by asking ChatGPT once. My tracker does three things.

  • It writes the questions. — Claude Code reads our website and a short company profile and writes the questions our buyers would ask. That gave us 53.
  • It asks every engine. — Each week it puts all 53 questions to ChatGPT, Claude, Gemini and Grok through their APIs.
  • It records the answers. — Who gets named, in what position, and which domains get cited. Because it runs weekly, I can compare this week with last week.

The first check is the most basic one: when someone asks an engine about OneAway by name, does it actually talk about us? You’d be surprised how many companies have a name and a domain, and an AI asked about them still answers about something else. We get named in all four engines, so that one’s fine.

Then there’s share of voice: how often you’re named compared with the competitors who show up for the same questions. We’re at the bottom of that list. I’m up against agencies that make hundreds of millions, so I expected that, and we’ll keep climbing.

There’s also a list of top cited domains. Read it with care. Some of the questions mention OneAway by name, and an engine will cite our site for those, so our own domain looks better there than it really is.


Read your rank on every buyer question

The view that matters most is the one with the questions. For each question, it shows where we rank on each engine.

The one I care about most is “best cold email agency for B2B SaaS”. That’s the work I want, and it’s what we are. Today we rank number two on Grok and number six on Claude. On ChatGPT and Gemini, we’re named but far down the list.

How to read the view:

  • A number — Your position in that engine’s answer.
  • A coloured dot with no number — You’re named, but low down.
  • An empty dot — You’re not named at all.
  • The trend — Whether you moved up or down on that question since last week.

You can edit the questions, or refresh them and let the AI write new ones. It keeps the ones it’s most sure your buyers ask. Be honest about quality, though. Some of mine aren’t good. The best source is your customers. Ask them what they searched for before they bought from you, or before they booked a meeting with you, and put those questions in.


Action 1: write a page

Measuring is half of it. The jump on Grok came from what I did next. Every weekly run produces a lot of data, so I told Claude Code to turn what it found into a list of actions. Each action is one of four types. The first is writing a page.

A write-a-page action points at a page the engines cite a lot and tells you to publish your own version. One example: “Publish your own version of cleverly.co’s page.” The engines had cited it in 22 answers to 13 buyer questions. That’s the kind of page you want to own.

Each action has a copywriting prompt. Click it and you get a full brief to hand an AI agent so it can build that page on your website. To make my life easier, I gave Claude Code access to the tracker and told it to work through every write-a-page action. I think it’s the lowest-hanging fruit, and most of them are now done. The cleverly.co action became our guide to cold email agencies with a live client portal.


The second type is getting onto the sites the engines already cite. The tracker first checks whether the cited site belongs to a competitor. Asking a competitor “You’re cited by AI a lot, can you put me on your page?” goes nowhere, so it skips those.

What it looks for is vendors: the software companies in your space. That makes a lot of sense for an agency like ours. Salesforge, for example, now shows up all over AI answers in our space, and I want to be on its pages.

Because we’re a cold email agency at heart, this action ends in outreach. Click “Find people” and it finds the people most likely to decide which agencies a site features. Pick the ones you want and it finds their emails. The card moves to in progress, and the leads go into a list called AI outreach, with a column for the article we want to be on. From there it’s one step to add them to a campaign. Parts of this are still in beta.


Action 3: fix a fact

The third type is about wrong answers. You don’t want an engine confused about you and repeating it to buyers.

For us, it found two things. One was an old product name I no longer use. The bigger one was pricing: after we changed our prices, old figures were still out there, and pages disagreed with each other. I fixed that. If you’ve ever changed your pricing, your positioning or your product names, check this first.


Action 4: join the conversation

The fourth type points at threads the engines cite, such as a Reddit post that comes up in a lot of answers. The idea is to reply there and mention OneAway where it’s relevant, so the next time an engine uses that source, we’re in it.

Some of these are easy wins. One action pointed me to a YouTube video the engines cite that had no comments at all. Nobody to compete with. I can leave one long, useful comment, and the next time an AI cites that video, my name is there.


Bonus: get your new pages indexed

If you want results as fast as I got them, make sure Google indexes every page you write or edit. Log in to Google Search Console at search.google.com, inspect each new or changed URL and request indexing. You can tell your agent to do it, or do it yourself. It helps both Google and the AI engines find the page sooner.

Why bother? My understanding is that it matters most for smaller sites. If Apple publishes a page, it gets picked up right away. A site with a modest domain rating, like mine and probably yours, can wait weeks.

Two limits, from Google’s own documentation on asking Google to recrawl a site, checked in November 2026:

  • There’s a quota. — Google says there’s a quota for submitting individual URLs but doesn’t publish the number. In my account I can’t request more than 10 pages manually. Requesting the same URL twice doesn’t make it faster.
  • Sitemaps for volume. — For a large number of URLs, Google says to submit a sitemap instead. It also says crawling can take from a few days to a few weeks, and a request doesn’t guarantee the page gets indexed.

Frequently asked questions

How long does it take to get recommended by AI?

Our Grok number went from 7.1% to 42.9% between two weekly runs, on October 24 and November 3. That doesn’t mean it can be hacked overnight. We still rank low on ChatGPT and Gemini, and I’ll be watching how it holds up over the next few months.

Which AI engines should I track?

I track ChatGPT, Claude, Gemini and Grok. Track them separately. For the same question, we rank number two on Grok, number six on Claude and much lower on the other two.

Where do the questions come from?

Start by having an AI write them from your website and company profile. Then improve them with your customers’ own words: ask what they searched for before they bought or booked a meeting with you.

Do I need a paid AEO tool?

Not to start. I built my tracker with Claude Code in about a week, and I keep adding to it. If you’d rather buy one, see our ranking of answer engine optimization tools.

Should I ask competitors to feature me on their pages?

No. They have no reason to say yes. Look for cited sites that belong to vendors or partners in your space, and reach out to the person who decides what those pages feature.

An AI engine says something wrong about my company. What do I do?

Find where the wrong fact lives and remove or correct it, especially old pricing and retired product names. Engines repeat what they find, so pages that disagree with each other confuse them.

Why request indexing in Google Search Console?

So Google crawls a new or edited page sooner, instead of waiting until it finds it. Google limits how many individual URLs you can submit, so save requests for the pages that matter and use a sitemap for the rest.


Key takeaways

  • Measure first: ask ChatGPT, Claude, Gemini and Grok your buyers’ questions every week and record who gets named and what gets cited.
  • The question view matters most. Your rank on each question, on each engine, tells you where to work.
  • Turn the data into four kinds of action: write a page, get featured, fix a fact and join the conversation.
  • Writing pages is the lowest-hanging fruit. Get featured on vendors’ sites, not competitors’.
  • Request indexing for every new or edited page in Google Search Console, within Google’s quota.

Want outbound run by AI agents we built?

We run cold email and LinkedIn outbound for B2B companies, from lead lists and data validation to copywriting and CRM sync, and book qualified meetings onto your calendar. We also work on AEO, so AI answers name your business. Junction grew from around $3M to $20M ARR during the almost three years we ran their outbound. Read the Junction case study.

Book a call