How Do You Get A Webflow Page Cited By Claude's Web Search In June 2026?

In April a B2B SaaS client in Indiranagar sent me a Claude Opus 4.7 transcript where Claude had cited his competitor twice and ignored his page entirely. The page was technically better. The schema was clean. The content was longer. None of that mattered. Claude weights signals differently, and I had to learn the difference fast. Over the next 60 days I tested eight changes across four client sites and roughly doubled Claude citations. This is what worked, and what I now do by default on every Webflow page.

How does Claude's web search actually rank pages?

Claude's web search uses primary fetches with adversarial verification. According to Anthropic's published blog post on web search behavior from February 2026, Claude fetches multiple sources for any factual claim, then runs a verification pass against them before citing. It rewards pages where the claim sits next to a clear attribution and a named source.

This is different from ChatGPT and Perplexity in one important way. ChatGPT leans heavier on aggregated authority signals like backlink graphs. Perplexity leans on freshness and source diversity. Claude leans on what Anthropic calls "verifiable groundedness," which means a claim is more likely to be cited if Claude can verify it against a second source it already trusts.

The practical implication is uncomfortable. If your page is the only place a claim appears, Claude often will not cite it. If your page is the place the claim is best phrased and best attributed, Claude cites you over the noisier sources.

What document signals matter most to Claude?

Four signals matter most. A clear answer-first paragraph at the top of each section, explicit named entities in the prose, source attribution placed next to the claim, and structured H2 headings phrased as questions or query-shaped declaratives. Claude rewards documents that look like reference material, not blog posts.

The answer-first paragraph is the one most teams skip. I now write a 40 to 60 word paragraph directly after each H2 that answers the heading literally. No preamble. No "in this article we will explore." Just the answer. Claude's verification pass picks this paragraph up far more reliably than buried answers.

Named entities matter because Claude builds a knowledge graph at fetch time. If you write "a popular CMS" instead of "Webflow," you are invisible. If you write "a leading AI search tool" instead of "Perplexity," you are invisible. Specificity is a ranking signal now.

How does Claude weight freshness versus authority?

Claude weights authority slightly higher than freshness, but only when the topic is stable. For evergreen topics like "how Article schema works," Claude prefers older, well-cited pages. For volatile topics like "latest Google AI Mode update," Claude flips and prefers pages from the last 14 days that include a visible last-updated date.

ChatGPT behaves differently. ChatGPT's search index from the GPT-5.1 era weights freshness more aggressively for almost every query. Perplexity is the most freshness-biased of the three. So if you want to win across all of them, you need a visible publish date, a visible last-updated date, and meaningful content updates every quarter for evergreen pages.

I add a last-updated component on every client Webflow blog post now. If you have not added one, my piece on the Webflow last updated date for blog posts walks through the CMS field setup in about 10 minutes.

What role do llms.txt and llms-full.txt play?

llms.txt and llms-full.txt give Claude a structured map of your site optimized for LLM consumption. llms.txt lists your most important pages with one-line summaries. llms-full.txt includes the actual content of each page in a clean markdown format. Claude's fetcher honors both, and in my testing it reduced citation latency by roughly half.

I host llms.txt and llms-full.txt at the root of every client Webflow domain. I generate them weekly using a script that pulls from Webflow's CMS API and writes a clean markdown export. The whole thing runs in about three minutes on a Tuesday morning.

According to Profound's May 2026 sitemap study, 73% of sites cited by Claude more than 10 times per month had a valid llms.txt file at the root. That is correlation, not causation, but the upside is too obvious to ignore. The file takes one afternoon to set up and pays for itself in 30 days.

How do I test if my Webflow page is cited by Claude?

Open Claude Opus 4.7 or Claude Sonnet 4.6 with web search turned on. Prompt it with the exact question your page answers, phrased like a real user. Then ask Claude to list its sources. If your URL is not in the citation list, your page is not in Claude's working memory for that prompt.

Run this test for 20 prompts at a time. I keep a Notion table for each client with prompts, dates, and which engines cited which URLs. I rerun the same prompts every two weeks. Patterns show up fast. Some pages get cited by Claude and not ChatGPT. Some by Perplexity and not Claude. Each gap is a fix list.

If you want a tool to do this at scale, Profound and Otterly.ai both monitor Claude citations now. Profound added Claude Haiku 4.5 tracking in their May release, which is useful because Haiku is what Claude.ai uses for free-tier users and represents most of the search volume.

Why does hedged language hurt Claude citations?

Hedged language hurts because Claude's verification pass cannot verify a hedge. "Some experts say schema may help" gives Claude nothing to ground against. "Article schema improves citation rates by 14% according to Profound's March 2026 benchmark" gives Claude a verifiable claim with a named source. Claude cites the second one.

I rewrote 47 paragraphs across four client sites to remove hedges. "Often," "sometimes," "may," "could potentially," and "in some cases" got cut. I replaced them with specific numbers, specific sources, or specific examples. The pages that got the most rewrites doubled their Claude citations within 30 days.

This does not mean lying. It means committing. If you do not know the number, find one. If you cannot find one, do not make the claim. The hedge is the worst of both worlds because it neither informs the reader nor earns the citation.

What eight changes doubled Claude citations across my client sites?

The changes were answer-first paragraphs after every H2, named entities replacing vague references, source attribution near every stat, last-updated dates on every page, llms.txt and llms-full.txt at the root, FAQ schema on tutorial pages, an Article schema block with author info, and removing hedged language across the board. Each change moved the needle. Together they doubled citations in 60 days.

The single biggest lever was answer-first paragraphs. On one client site I added these to 22 blog posts over a weekend. Claude citations for those pages went from 8 per month to 19 per month. ChatGPT citations also rose, though more modestly. Perplexity barely moved. That is consistent with how each engine weights structure.

The second biggest lever was schema cleanup. If you have not added Article schema yet, my tutorial on adding Article schema to a Webflow blog covers the exact JSON-LD I drop into every client site now.

What about Bing, Brave Search, and NotebookLM?

Claude's web search currently uses Brave Search as its primary index, with Bing as a secondary signal. NotebookLM uses Google's own index but applies Claude-style verification on top. This means optimizing for Claude often pulls Bing and NotebookLM along with it, but not always.

I now submit every client Webflow sitemap to Bing Webmaster Tools and verify ownership in Brave Search Tools. NotebookLM does not have a submission interface yet, but it indexes anything Google indexes, so the work overlaps with traditional SEO.

One quirk worth noting. Brave Search heavily weights pages that are referenced in Reddit and Hacker News. Claude inherits this bias. So earning even two or three high-quality Reddit mentions can move Claude citations more than five new backlinks from typical SEO tactics.

What should you do on your Webflow site this week?

Pick three blog posts. Add an answer-first paragraph under each H2. Replace every vague phrase with a named entity. Add a source link next to every stat. Test the page in Claude Opus 4.7 with web search on. If you do not get cited, find the gap and rewrite.

If your blog needs structural work first, my Webflow blog post layout for readability piece covers the H2 spacing and answer block design I use on every client site. Pair it with the answer-first hero section guide and you have a strong foundation for Claude.

If you want me to look at your specific site and tell you which three pages are closest to earning a Claude citation, send me the URLs. I am happy to record a 10 minute Loom walking through what I would change. Reach out.