How many software company websites have the FAQs, tables and question headings AI engines quote?

Of the 1,303 software company websites where my check found article pages, 51.9% had an FAQ section, 33.4% had a comparison table with header cells and 21.3% passed the question-heading check. Only 53 of those sites (4.1%) passed all four answer formats I test for, and 363 (27.9%) passed none. The figures come from 1,601 software company websites I put through version 2 of my AI search readiness check on 22 October 2026, of which 1,586 were scored. Another 283 sites (17.8% of the 1,586) had no article page the check could find, and they averaged 32.1 out of 100 against 70.8 for the sites with articles. Two of the four formats, FAQ and comparison tables, also predicted a citation in my 209-vendor Perplexity test.

By Ronan Liedmeier, fractional head of growth. Last updated 23 October 2026.

An answer format is a piece of page structure an engine can lift whole and quote: a question with its answer, a table row, a heading phrased as the buyer's question with a short answer under it. The check tests four of them (question headings, answer-first sections, comparison tables, FAQ) plus the evidence around them: depth, facts, sources, dates and authors.

This article is one part of the 1,586-site readiness study. The score estimates whether a page can be read and quoted; it counts no citations.

Which formats did the check look for, and how did software sites do?

The check tests the four answer formats plus content inventory, depth and the evidence checks (facts, sources, dates, recency, authors). Visible dates (61.6%) and depth (56.9%) pass on most article sites, while question headings (21.3%) and outbound sources (23.6%) pass on fewer than one in four.

Every figure below is a share of the 1,303 sites where the check found article pages.

Tables and FAQ have no partial level. B1 cannot fail on this base.

The check reads raw HTML. Vercel's crawler analysis found that "none of the major AI crawlers currently render JavaScript" (Vercel), so an FAQ accordion or table that only appears after a script runs counts as missing (more in the JavaScript article).

How many software sites have no article pages at all?

283 of the 1,586 scored software company websites (17.8%) had no article page the check could find. Their mean score was 32.1 (median 36), against 70.8 (median 72) for the 1,303 sites with article pages.

Part of that gap is built into the scoring. When no article page is found, the question-heading, answer-first, table and FAQ checks fail automatically and depth is marked not applicable. For these companies the first job is a small set of pages that answer buyer questions, before any format work.

How common are FAQ sections on software websites?

51.9% of the 1,303 article sites had an FAQ heading or FAQPage markup on the homepage or a sampled page. It is the most common of the four formats and the only one that passes on more than half of the sites.

FAQ was the strongest predictor in the 209-vendor test, tied with named authors: within the same Perplexity answer, the vendors whose site was cited had an FAQ section more often than the vendors the answer named without citing.

Google no longer shows FAQ rich results. It added a deprecation notice to its FAQ rich result documentation on 8 May 2026, saying the feature "will no longer appear in Google Search starting May 7, 2026", and removed the documentation on 15 June (Google Search Central). The question-and-answer text is still on the page for an engine to lift.

How many software sites use comparison tables?

33.4% of the 1,303 article sites had at least one HTML table with header cells on a sampled page. 66.6% had none, the highest fail share of any check in the table above.

The threshold is low: one table on one sampled page passes. It has to be an HTML table with header cells, so a comparison drawn as an image or a grid of divs does not count, because neither carries row and column labels an engine can read as data.

Comparison tables predicted citation within an answer in the 209-vendor test.

Only 338 sites (25.9% of the 1,303) had both an FAQ and a table, the two answer formats in this set that predicted citation.

Do software sites write question headings and answer-first sections?

Mostly not. 21.3% of the 1,303 article sites passed the question-heading check (30% or more of H2/H3 headings phrased as questions) and 30.2% passed the answer-first check (60% or more of H2 sections opening with a 15 to 90 word paragraph). The fail shares were 35.3% and 40.1%.

These two checks cost more points than any other check in the study: across all 1,586 scored sites, question headings lost 6,156 points and answer-first sections 5,994.

The citation evidence is weaker than the point loss suggests. Answer-first paragraphs showed no citation effect in the 209-vendor test, and that write-up reports no result for question headings. So I treat both as cheap structure fixes with no proven citation effect yet. Rewriting an H2 as the buyer's question with a two-sentence answer under it is a quick edit that makes a page easier to scan.

Question headings showed no US and Europe gap: 21.9% against 20.9%.

How deep, factual and sourced are software company articles?

Depth is in reasonable shape and sourcing is not. 56.9% of the 1,303 article sites had a median of 800 or more words per article page, while only 23.6% had a median of three or more outbound links inside article content. On facts, 33.0% reached 8 or more numbers, percentages, prices or dates per 1,000 words of article text.

Depth predicted citation within an answer in the 209-vendor test. Outbound sources were weak there and did not reach significance. Facts per 1,000 words is not in that test's published table.

43.6% of article sites failed the sourcing check, meaning their median article had fewer than one link to another site in the body. A linked source makes a claim easier to check. The citation effect is unproven, so I put it after the five checks that did predict.

Do software articles show dates and named authors?

61.6% of the 1,303 article sites showed a date or dateModified on 60% or more of their article pages, and 49.8% passed the named-author check. Both predicted citation within an answer in the 209-vendor test.

The author check needs two things: 60% or more of article pages name an author, and an about or author page exists. 40.4% of sites met one of the two: names without an about or author page, or the page without names on the articles. Google's guidance asks the same question: "Do pages carry a byline, where one might be expected?" (Google Search Central).

Dates and recency are different checks, and only one predicted citation. Visible dates (C3) asks whether the reader can see when the article was published or updated. Recency (C4) asks whether the newest date found, sitemap lastmod included, is within 90 days. 80.7% of article sites passed recency, and it showed no citation effect in the 209-vendor test. The date a reader can see on the article is the one that mattered in that test.

How many software sites have all four answer formats?

53 of the 1,303 article sites (4.1%) passed all four formats: question headings, answer-first sections, a comparison table and an FAQ. 363 (27.9%) passed none, and the average site passed 1.37.

The rising mean score is partly mechanical, because each format is worth points. The useful reading is how rare the full set is.

Do US and European software sites differ on formats?

On two formats, yes. US article sites had an FAQ more often (62.3% against 45.0%) and a comparison table more often (42.7% against 27.2%), and both gaps are large enough to be real.

Europe is the 12 European countries in the sample (947 scored sites; US 639). The overall score is about the same, and the difference sits in two formats that predicted citation. I do not rank individual countries; most have too few sites.

Which format gaps matter most?

The gaps worth closing first are comparison tables and FAQ sections: they are missing on 66.6% and 48.1% of the 1,303 article sites, and both predicted citation within an answer in the 209-vendor test. Dates, depth and named authors come next, because they predicted citation too and are missing less often.

I put named authors fifth despite the lowest fail share, because 40.4% of sites sit at partial and the fix (an author page behind the name) is quick. The 209-vendor results come from one engine and one run, with modest correlations, so treat this order as provisional.

Method

• Sample: software publishers (NAICS 51321 or 5112), 11 to 200 employees, founded 2015 or later, headquartered in 13 countries, pulled from Apollo on 22 October 2026. Agencies, consultancies, PR firms, online shops and staffing firms were excluded by NAICS code. 32 companies were removed before the run: 12 I had already contacted and 20 very large well-known brands. The flagship study has the full exclusion list.
• Run: one continuous batch on 22 October 2026, 18:03 to 18:28 Pacific time, with version 2 of the same scoring code as the free tool, run locally on a residential connection. One check per site, median 9 URLs read.
• Results: 1,601 attempted, 1,586 scored (3 of them partial scans that ran out of the 45-second budget), 15 failed (12 domains did not resolve, 3 other errors).
• Base for this article: the 1,303 sites (82.2%) where the check found article pages. The 283 sites without them fail the format checks automatically, so including them would mix "no articles" with "articles without the format".
• Limits: the check samples up to eight pages per site, so a table or FAQ on an unsampled page is missed. Thresholds are mine. Formats like FAQ and tables are tied to citation only through the 209-vendor test, and that test found no citation effect for Organization schema, sameAs, answer-first paragraphs or recency. Ahrefs' controlled test of pages that added JSON-LD found no clear citation uplift either (Ahrefs). No company from the sample is named here.

What I'd do

I add formats to an existing article in this order, most proven first.

1. Add a comparison table as an HTML table with header cells: options against criteria, or plan against feature. It is the format missing most often and it predicted citation.
2. Add an FAQ block of three to five questions in visible text, each answered in two to four sentences with a number in it. The Google rich result is gone, so the text is what counts.
3. Put a visible published or updated date on the article itself, near the title. The recency check, which also reads sitemap lastmod, did not predict citation.
4. Name the author on the page and link the name to an about or author page with their role and background. If you have names without pages, this is the quickest pass in the list.
5. Bring thin articles past 800 words of median length by answering the follow-up questions a buyer asks next.
6. Link two or three outside sources inside the body where you make a claim, and add the numbers you already know: prices, dates, percentages. Weak or untested for citation, cheap to do.
7. Last, the structure pass: rewrite H2s as the buyer's questions and open each section with a 15 to 90 word answer. No proven citation effect yet, and a quick edit per article.

For the full audit list behind these checks, see the 12-point GEO audit checklist, and for everything beyond formats, the guide to getting cited by ChatGPT, Perplexity and AI Overviews.

I do this work as a fractional head of growth for software companies that want their articles to be the source an AI answer quotes. The first step is the free readiness check: it runs the same 22 checks on your site and shows which of the formats above your articles are missing. If you want the citation side measured too, the AI Visibility Snapshot is ten prompts in your category, your site scored on the 22 checks, and a one-page readout in three working days.

Re-check by 23 January 2027: re-run the format checks on the same 1,303 article sites to see whether FAQ and table shares move, whether the European gap on FAQ and tables narrows, and whether the second 209-vendor run on ChatGPT and Google AI Overviews confirms which formats predict citation and gives a result for question headings.