Every agency on the internet is selling answer engine optimization right now, and the advice is remarkably consistent. Add FAQ schema. Put the answer in the first sentence. Write longer. Use lists. Match your heading to the question. We wanted to know whether any of that actually separates a page that wins from a page that loses, so we measured it.
It mostly does not. One thing does, and it is the oldest tactic in search.
How We Ran This
The hard part of studying answer placement is that the winners are usually big sites, and big sites win for reasons that have nothing to do with page structure. Comparing a Google support page to a plumber's blog post tells you about authority, not about tactics.
So we built the comparison to cancel that out. For each of 17 question shaped queries we took the pages Google ranked in the top 3 and compared them against the pages Google ranked 7th to 10th for the exact same query. Both groups cleared the same relevance bar. Both are about the same thing. Anything that differs between them is a difference in how the page was built, not in what it was about.
Then we fetched all 76 pages and measured the classic on page signals on each one. Word count. Heading count. Whether a heading matched the question. Whether an answer sat directly under that heading. Schema types. Lists. Tables. A number in the opening. Internal links. Outbound links.
Our first run said winners were shorter, lighter on schema and thinner on links than losers, which would have been a fun headline and a wrong one. The top 3 group was 60 percent Reddit threads and Google's own documentation, and those pages are naturally short and plain. We stripped every platform out and re ran it against independent business sites only. The numbers below are from that second pass, 18 winning pages against 22 losing ones.
More Than Half the Answer Real Estate Is Not Yours
Before any tactic matters, there is a ceiling worth knowing about. Across our 17 queries there were 42 positions in the top 3. 24 of them went to Reddit or to a Google property. Reddit alone took 9. Google support, Google Business and Google's developer docs took 15 between them.
That is 57 percent of the answer real estate on question queries sitting somewhere your website cannot go. It does not mean the other 43 percent is not worth winning. It does mean that anyone promising you the answer spot on a question query is promising something they only control about four times in ten.
The AEO Checklist Did Not Separate Winners From Losers
Here is every signal we measured, winners against losers, on the same queries. Medians for counts and averages for the yes or no signals.
| What we measured | Top 3 | Ranks 7 to 10 | Difference |
|---|---|---|---|
| Word count | 2,364 | 2,297 | 1.03x |
| Distinct schema types | 10 | 10 | 1.00x |
| Has FAQ schema | 22% | 18% | 1.22x |
| Has Article schema | 67% | 77% | 0.86x |
| Lists on the page | 22.5 | 21 | 1.07x |
| A number in the first 250 words | 83% | 82% | 1.02x |
| Question words in the title | 0.75 | 0.77 | 0.97x |
| Headings on the page | 28 | 23 | 1.22x |
| Answer sits under a matching heading | 89% | 77% | 1.15x |
| Internal links in the content | 29 | 12 | 2.42x |
| Outbound links to sources | 9 | 2.5 | 3.60x |
Read down that table and the first nine rows are noise. A 1.03x difference in word count across a sample this size is nothing. Schema volume was identical. FAQ markup showed up on four winners and four losers, which at this sample size is a coin flip wearing a lab coat. Having a number early, matching the question in your title, writing more lists, none of it told the two groups apart.
The two rows at the bottom did.
Internal Linking Was the Signal
The typical page in the winning group carried 29 internal links inside its content region. The typical page ranked 7th to 10th carried 12. Same query, same intent, two and a half times the internal linking.
We counted those links properly, which matters more than it sounds. Our first pass only counted
links written as /like-this and missed every site that writes internal links as full
URLs. Recounting by resolving every link against the page it sits on moved the difference from 1.7x
to 2.4x. If you run this kind of audit yourself, resolve the host. Do not pattern match the href.
The average internal link count was almost identical between the two groups, 39.3 against 40.0. Only the median moved. That means a handful of low ranking pages were carrying enormous link counts and dragging their own average up. More internal links is not better without limit. The pattern is that winners cluster around a healthy number and losers are split between pages with almost none and pages that are stuffed.
The second signal is outbound linking. Winning pages cited other sites a median of 9 times against 2.5 for the losing group. Linking out to your sources is advice from 2004. It still tracks with winning the answer spot in 2026, and almost no AEO checklist mentions it because it sends people away from your page.
Why This Result Makes Sense
An answer engine has to decide which page on a topic is the authoritative one. Schema tells it what your page says. Links tell it what your site knows.
A page with 29 internal links is usually a page sitting inside a real body of related work, with somewhere to send a reader who wants the next thing. A page with 12 is more often a standalone post written to rank, with nothing around it. The link count is not the cause. It is the fingerprint of a site that covers its subject properly, and that is what the engine is trying to find.
This is also why the checklist items came out flat. Everyone already does them. When every page on the first two results pages has FAQ schema and a number in the opening paragraph, those things stop being a way to tell pages apart. They are table stakes, not an edge.
What To Do With This
- Build the cluster before you optimize the page. One article about a topic will not beat a site that has covered the topic ten ways and links it all together. The link count is a symptom of that work, so chasing the number without doing the work will not help.
- Link out to your sources. It was the strongest single ratio in the study at 3.6x, and it is free.
- Keep the schema, stop expecting it to win. Structured data helps machines parse your page and it is cheap to add. Our data says it will not lift you past a competitor who also has it.
- Check whether the slot is even winnable. If Reddit and Google's own docs hold the top 3 on your query, your effort belongs on a different query.
- Stop paying extra for AEO as a separate service. On this evidence it is local SEO fundamentals with a new name on the invoice.
What This Study Cannot Tell You
We would rather say this out loud than have you find it.
- The sample is small. 18 winning pages against 22 losing ones, across 17 queries, after platforms were removed. It is enough to see a 2.4x gap. It is not enough to trust a 1.2x gap, which is why we are not making anything of the heading counts.
- This is correlation. We did not add internal links to pages and watch them move. Pages that rank well may simply live on better sites, and better sites link more.
- It is one snapshot. United States results, desktop, gathered on a single day in September 2026. SERPs move.
- Content region detection worked on 31 of 40 pages. On the other 9 we counted links across the whole page, which inflates both groups equally but adds noise.
- We could not measure AI Overviews directly. Google does not expose the Overview block to the tooling we used. We measured the organic answer pool instead, which is where Overviews draw from. Every one of the 68 People Also Ask entries we captured was flagged by Google as an AI Overview type question, so the overlap is real, but it is a proxy and we are calling it one.
Final Word
Answer engine optimization is being sold as a new discipline. Our numbers say it is the old one with better marketing. The pages winning answers are not the pages with the most schema. They are the pages that sit inside a properly built site and are honest about where their information came from.
That was the job in 2016 and it is the job now. If you want to see how we do it, our local SEO services page has the pricing and the guarantee, and our guide to choosing an agency has the questions to ask whoever you hire.