We measured our answer blocks. The honest result was small.

Adding answer blocks to pages that already rank on the second page of Google moved them up by a small amount on average, and about two thirds of the pages improved. Our first tally of the same work said the pages gained almost five positions. The gap between those two figures is the useful part of this post, because the mistakes that inflated our number are easy to make in any before and after check.

What an answer block is

An answer block is a section written to answer one search directly. It opens with a heading close to how people ask the question, gives a one or two sentence answer straight away, and follows with two or three lines of support. The claim and the evidence sit in the same chunk, so the block still makes sense if a search engine or an AI tool lifts it out of the page.

We started adding them on 4 August 2026. The first batch was 32 pages across six of our sites, chosen because Google already showed each page between positions 11 and 30. Pages in that band move faster than new ones, and the page's own search data shows what it fails to answer. More batches followed through 24 August, 97 pages in all in our log.

The first result, and why it was wrong

Our log said the median page gained 4.8 positions and 21 of 30 measured pages went up. That looked excellent. On 1 September we re-measured with stricter rules and the picture changed:

• Compare the same searches. A page's average position moves whenever the mix of searches behind it changes, even if nothing improves. We compared only searches that appeared in both windows, weighted by how many impressions they had before.
• Subtract a control. Every site was drifting upward anyway. Across August the untouched pages on our sites gained between 1.2 and 3.4 positions on their own. A raw gain that size says more about the site than about the block.
• Only count crawled pages. We asked Google, page by page, when it last crawled each URL. About 25 of the 97 had not been recrawled since their block shipped. Their movement measured the old page.
• Merge address variants. The http, https, www and bare versions of an address were being counted separately.

With those rules the median gain was about 0.77 positions against the control at 28 days, and about two thirds of the pages improved. That is a real effect and a small one.

Three ways a result can mislead you

A collapse can look like a win. A page's average position is weighted by impressions. If a page loses hundreds of searches at positions 40 to 90 and keeps a few near the top, its average improves while its traffic falls. On a builder's site, two pages in our first batch fell sharply in impressions. Daily data showed the drop began on 1 August, three days before we shipped the blocks. A page in the same commit roughly doubled. The blocks were innocent, and the pattern would have looked like either a failure or a success depending on which number we read first.

Your own pages can compete. On a property site, a guide in the batch appeared to lose 29 positions. The cause was a sibling post published nine days earlier on a near-identical topic, not the block. Check whether another page of yours now answers the same search.

Your control can be contaminated. When we re-measured one site's batch on 30 September, the default control read plus 6.0 positions. A clean control, excluding every page that had been edited for any other reason in the same window, read plus 2.3. The median result for that batch moved from 0.08 to 1.29.

What about AI answers

One of our pages, a fridge repair post on the same appliance site, started appearing for "how to fix refrigerator" at position about 1 with zero clicks and a jump in impressions. That pattern looks like a citation in an AI Overview, where the page is quoted and the searcher does not need to click. Meanwhile the page's ordinary results drifted down, which made it look like a regression in the numbers.

We would not read much into one page. What it does show is that a visible win in an AI answer can arrive as lost clicks, so judging answer blocks only by traffic undersells them, and judging them by position alone gets the picture wrong. Our post on whether your business shows up in ChatGPT covers the broader question.

What to do with this

1. Add answer blocks to pages you already rank for between positions 11 and 30, using the real searches in Search Console as your headings.
2. Wait until Google has recrawled the page before measuring anything. Our post on sitemap dates and recrawling explains why that can take longer than you expect.
3. Look at the searches behind a page, not its average, as in our post on what average position hides.
4. Compare against pages you did not touch, and make sure those pages stayed untouched.

Treat the effect as small and worth having. A block costs little and sits on a page that is already close, which is a better use of time than a new post, but it will not rescue a page that is nowhere near page one.

Where this sits with our plans

Choosing the pages, writing the blocks and checking the result honestly is the sort of work the SEO add-on covers. Prices are on the pricing page, and you can get started any time.