The AEO Checklist: 11 Fixes That Decide Whether AI Names Your Company
In July 2025, Ahrefs reviewed 1.9 million AI Overview citations. It found that 76% came from pages sitting in Google's top 10. In March 2026, it ran the numbers again using 863,000 keyword searches. This time, the share was 38%.
Eight months. The link between rankings and getting named had been cut in half.
Most companies still put nearly all their content money into ranking. This page focuses on getting named. It covers the 11-point check we use on a founder brand or B2B site before doing anything else. Every point comes with its data. Score your own site before spending a rupee.
One fair warning first. Ahrefs says the two studies are not a perfect match. Its citation tracking got better between them, so some of the drop comes from better measurement. The wider direction still stands. Google's own documentation also explains why this is happening. We will cover that below.
What is an AEO checklist?
An AEO checklist is a fixed set of page and brand fixes that shape whether AI assistants name your company when a buyer asks who to hire. Moxie's version has 11 points, and each one comes with its source.
AEO means answer engine optimization. The goal is to get your company named and quoted inside AI answers, beyond appearing on a search results page.
This matters because the buying process has changed. Gartner surveyed 646 B2B buyers between August and September 2025 and published the findings on 9 March 2026. 67% said they prefer buying with no sales rep at all. 45% used AI during a recent purchase. Your next client could first see your company as one line inside a machine's answer. This checklist shapes what that line says and whether your name appears.
Keep score while you go. Give yourself one point for each item you already do. Write down the total. Most founders score 3 or 4 out of 11 on the first check. That is common.
Why does ranking on Google no longer decide whether AI names you?
AI Overviews now take most citations from pages outside the top 10 results. Ahrefs found that top-10 pages supplied 76% of cited pages in July 2025. By March 2026, the number was 38%.
Here is the full view from the March 2026 study covering 4 million AI Overview URLs:
| Where cited pages rank in Google | July 2025 | March 2026 |
|---|---|---|
| Top 10 | 76.1% | 37.9% |
| Positions 11 to 100 | 9.5% | 31.2% |
| Beyond the top 100 | 14.4% | 31.0% |
Source: Ahrefs, study of 1.9M citations, July 2025 and study of 863K keywords, March 2026. Ahrefs says its parsing got better between the studies, so these numbers show direction rather than a clean like-for-like comparison.
Look hard at the middle and bottom rows. Nearly a third of citations now come from pages sitting outside the top 100. A page that gets almost no normal search visibility can still become the source an AI answer quotes.
What drives this? Google's own search documentation says its AI features use a process called query fan-out. A buyer enters one question. Google breaks that question into several smaller and more exact questions in the background. It then pulls citations from pages that answer those smaller questions well, even when those pages have no meaningful ranking for the original phrase.
That changes how you should write. One page aimed at one phrase is no longer enough. Build the page around the smaller questions sitting inside the larger one.
Another result from the same study deserves attention. Of cited pages outside Google's top 100, 18.2% were YouTube URLs. Ahrefs also says YouTube is now the most cited domain in AI Overviews. If you have recorded a clear explanation of what your company does, that video may already be doing more work than your homepage.
Where on a page do AI engines take their answers from?
Near the top, most of the time. Kevin Indig's Growth Memo research from February 2026 identified 18,012 verified citations from 3 million ChatGPT answers. The study found that 44.2% of citations take material from the first 30% of a page.
Indig's team describes this shape as a ski ramp. Citation activity is highest near the start and drops as the page continues. The machine behaves a lot like a rushed buyer. It begins at the top, grabs the useful part, and often leaves before your sixth paragraph.
That makes the first checklist fix one of the easiest and cheapest. Give the answer inside the first 60 words of a section. Long warm-ups waste citation space. Opening with "before we answer this, some context" forces the useful answer too far down the page.
Try this on your own site. Open the service page that matters most. Read its first 60 words and stop. A stranger should be able to tell what you do, who you do it for, and what kind of price point you sit at. If those basics are missing, the page gives the year's most important reader very little to use.
Does schema markup get you cited?
Schema alone has shown almost no effect. Ahrefs tracked 1,885 pages that added schema markup between August 2025 and March 2026. It matched them against 4,000 similar control pages. Citation numbers barely changed across the AI platforms.
The detail matters because companies spend plenty of AEO money here. In the same May 2026 study, Ahrefs first reviewed 6 million URLs. Pages cited by AI were almost three times more likely to use schema than uncited pages. That looks convincing until you reach the experiment. Once pages added schema, Google AI Overview citations fell 4.6%. Google AI Mode and ChatGPT citations increased by amounts too small to mean anything.
The two findings fit together. High-quality pages often use schema, so cited pages show more schema too. Schema comes with page quality. It has not been shown to produce that quality or the citations by itself.
Our advice has stayed steady all year. Add schema. The cost is low. It gives machines cleaner page structure to parse and could matter more as engines change. Just treat it as basic plumbing. A vendor using the three-times correlation while leaving out the experiment is giving you half the evidence.
What are the 11 points of the AEO checklist?
The 11 points deal with two areas: how your pages are built and written, and how your company appears across the rest of the web. Here is the full list before we unpack each fix.
| # | The fix | What it fixes |
|---|---|---|
| 1 | Answer in the first 60 words | The machine reads the top of the page first |
| 2 | Write headings as buyer questions | Engines match the shape of the question |
| 3 | Stop relying on rankings alone | Top-10 pages now supply 38% of citations |
| 4 | Add schema, expect little from it | Parsing help, not a citation engine |
| 5 | Skip the llms.txt shortcut | No engine has committed to it |
| 6 | Use one name everywhere | Machines resolve entities by consistency |
| 7 | Put a number in every claim | Numbers are what machines quote |
| 8 | Date your pages | Engines prefer maintained content |
| 9 | Publish where models read | Third-party sources feed the answers |
| 10 | Give the model a person to trust | Attributed content beats anonymous content |
| 11 | Measure share of answer | Rankings no longer tell you the truth |
1. Answer the question in the first 60 words of the section. Growth Memo's February 2026 analysis found that 44.2% of ChatGPT citations come from the first 30% of a page. Give the answer early. Add the explanation after it.
2. Write headings as the question your buyer types. "How much does AP automation cost for a 200-person company" gives the engine more to match than "Pricing." Google's query fan-out turns broad questions into smaller ones. Question-based headings help your page line up with those smaller questions one at a time.
3. Stop assuming a top-10 ranking carries you. Ahrefs' March 2026 data shows top-10 pages providing 38% of AI Overview citations. Its July 2025 study put the number at 76%. Ranking still matters. Its grip on citation visibility has slipped.
4. Add structured data, and hold your expectations steady. Ahrefs' May 2026 experiment across 1,885 pages found that adding schema barely changed citation levels, even though cited pages are three times more likely to use it. Use schema to make parsing easier. Keep citation expectations low.
5. Skip the llms.txt shortcut. llms.txt is a proposed file that tells AI crawlers which content to read. No major engine has committed to using it. A paid llms.txt package buys you a text file.
6. Name yourself the same way everywhere. Use one company name, one founder name, one role line and one location across your website, LinkedIn, bylines and directories. Machines connect identities through repeated signals. Three different spellings split those signals across three weaker identities. We see this inside our own results: two unrelated companies use the word Moxie, and the domain keeps the answers separate.
7. Put a number in every claim you make. "We cut approval time" gives a machine little to quote. "We cut invoice approval from 9 days to 2" gives it something exact. The Princeton GEO study, run across 10,000 queries and published at KDD 2024, found that adding statistics, quotations and citations increased page visibility in AI answers by up to 40%. Machines can lift a clear number straight into an answer.
8. Say when you last checked. Put dates on pages and keep the figures current. Ahrefs' study of 17 million citations found that AI assistants prefer citing fresher content. A 2024 statistic sitting on an undated page looks like content that may have been left untouched.
9. Publish where the models already read. The March 2026 Ahrefs study found YouTube provides 18.2% of citations from pages outside Google's top 100. Engines also use community threads and industry lists. Your site is one source in a larger set. Some of the other sources can be easier and cheaper places to gain visibility.
10. Give the model a person to trust. Put a real name on the content. Add genuine credentials and a bio that explains why the author knows the subject. Machines check people alongside companies. Anonymous content enters that comparison with fewer identity signals.
11. Measure share of answer, not rankings. Write down 20 questions your buyer could ask. Test them on the AI engines every month and count the answers that name your company. Turn that count into a percentage. The next section shows the process, using tools you already have.
Which fixes come first?
Begin with the changes you can make in an afternoon and repeat across the whole site. The order by return is simple: answer within the first 60 words, use buyer questions as headings, keep one name everywhere, support claims with numbers, and show a visible date on every page.
Those five changes are writing habits. They do not need a large project. A founder with a small website can put them in place within a week.
Here is the first fix on a real page. A services page we reviewed this month started with: "In a world where technology moves faster every day, businesses need partners who understand the pace of change." That spends thirty words and gives away zero information. After the first 60 words, a machine still has no clue who the company serves or what it charges.
Now compare the rewrite: "We run finance operations for B2B companies with 50 to 200 staff. Typical engagement: month-end close cut from 9 days to 3, from 2 lakh a month." The word budget barely changes. The useful information transforms. A machine can pull out the company offer, buyer type, measured outcome and price band. That is fix one in practice. Repeat it in every section on every page tied to revenue.
The next group needs more time. Building third-party presence on the lists and platforms engines quote requires real placements. Strong author pages and credentials may need a design discussion. Schema takes a developer afternoon and helps with parsing.
Measurement can come last when you have never tracked this before. But if you have two hours free this month, consider starting there. It shows you which of the other ten fixes deserves attention first.
How do you measure AI visibility without buying a tool?
Create 20 questions a buyer could ask. Test them through the main AI engines each month. Count how many answers mention your company. Divide that number by the total. The result is your share of answer.
Here is the exact process we follow.
First, create the 20 questions. Make them buyer questions rather than questions about your company. Think "best answer engine optimization agency in India" or "who should I shortlist for founder-led B2B content." When your company appears only after someone searches its name, you have brand awareness. That gives you little evidence of pipeline visibility.
Second, test every question in ChatGPT, Perplexity, Gemini and Google's AI results. Open fresh sessions so previous chat history cannot tilt the answer toward you. Test each question three times because AI outputs can change from one run to another.
Third, score each answer in one of three ways. Named means your company appears by name. Cited means the engine uses your website as a source but leaves your company name out of the answer. Absent means your company does not appear. Also record the companies that were named instead. Those names point you toward the lists and pages where you need visibility.
Fourth, repeat the test next month with the same 20 questions. Focus on movement over time rather than the first score alone.
That first test is rough for plenty of founders. It is also the cleanest honest baseline we have right now. As the Ahrefs numbers above show, Google ranking positions have become much weaker predictors of AI visibility. We use this same process on our own name every month. Our first run showed us things normal analytics had missed.
The best output is often the list of companies named instead of yours. When an engine answers a commercial question, it builds that answer from a small group of trusted sources. Many of those sources are industry lists and comparison articles. Ask the engine for its sources. Track the pages it keeps using. That becomes your outreach list for the quarter.
One placement on a page the engine already trusts can move your score faster than ten fresh posts on your own blog. The engine has already chosen that source. Getting included lets your company benefit from trust that source already holds. This is why checklist point 9 carries more weight than it first appears. For commercial questions, these lists are often the pages engines keep quoting.
What is the difference between AEO, SEO and GEO?
SEO gets your page onto a search results page. AEO gets your company into the sentence an AI assistant quotes. GEO gets your company onto the shortlist when a buyer asks who to hire. The query types change. The useful pages change. The work changes too.
There is overlap between all three. That overlap has also made it easy to sell the same work three times with different labels. The checklist above helps across all three because the same basics matter: clear page structure, claims backed by numbers, and one consistent company identity. The acronym of the month matters less than those fundamentals.
For a deeper view of where human work still matters, read our piece on whether AI can replace your marketing agency. AI answers also draw from every place your company publishes, so our article on why B2B buyers switch suppliers covers what happens when those different surfaces say different things.
FAQ
Does llms.txt help you get cited by AI? Evidence that it improves citations has not been shown. It remains a proposed standard, and no major engine has committed to reading it. Put that time into the eleven fixes above instead. Start with the first 60 words of each page.
How long until AI engines notice changes to your site? Use vendor timelines as rough guidance. Some engines may spot structural updates within days. Others may take weeks. We have not yet verified exact timing windows using client data, and we will not publish a number we cannot support. The monthly tracking process above will catch the change either way.
Is AEO different from SEO? Yes. SEO helps a page rank on a search results screen. AEO helps your company get named and quoted inside an AI response. Ahrefs' data shows the overlap between those two outcomes is far smaller now than it was a year ago. That is why this checklist treats them as different jobs.
Does adding schema markup improve AI citations? Schema on its own showed almost no improvement. Ahrefs followed 1,885 pages that added schema and found citation levels barely changed. Cited pages use schema more often, but strong pages often contain both. Add schema for parsing at low cost, then spend your real effort on clear answers, hard numbers and consistent identity.
Score yourself, then decide
Add up your points from the table above. Most founders finish at 3 or 4 out of 11. Moving from that score to 11 takes a fortnight of focused work. Every fix is already on this page, along with the source behind it.
One final practical point. Engines keep changing, so this checklist must change with them. Every figure here was measured between July 2025 and May 2026, and every number names the study behind it. We review this page every quarter and replace figures when newer editions appear. That is point 8 being used on the page in front of you.
If you want to see what AI engines say about your company today before making changes, book a call. We will test your name live using the same 20-question process above. You will leave knowing your actual number. Bring your score.