HubSpot SEO in 2026: The Content Hub Playbook for Search and AI Visibility

You came to this page because something about HubSpot SEO in 2026 isn't adding up.

Maybe your blog traffic is down 20–30% year over year and your CMO wants to know why. Or you're staring at HubSpot's new AEO product and wondering whether to buy it.

Maybe you're rebuilding your HubSpot site and want to know what good actually looks like in 2026 — past the recycled checklists that keep showing up on page one.

This guide is written for you.

Key takeaways

1. HubSpot SEO is a dual-track discipline — one playbook, two output formats
2. Most AEO work is SEO work at a higher standard — only passage chunking and llms.txt are genuinely AEO-only
3. Classic search = volume engine, AI search = conversion-rate engine — the 89% / 6% / 4.4–5× framing
4. Technical foundations matter more, not less — INP replaced FID, AI engines hit 499 errors on slow pages
5. HubSpot's native toolkit has changed — auto schema, Breeze Agents, the new $50/mo AEO product (Claude absent)
6. llms.txt has a HubSpot workaround — 30-minute File Manager + 301 redirect
7. Most AI traffic is misclassified as Direct — 70.6% arrives without referrer headers

It's the HubSpot SEO playbook we run for B2B clients on Content Hub in 2026. It covers what HubSpot's native SEO toolkit does (and where it lies), what's actually happening in Google's AI Overviews and AI Mode, how to use Content Hub's new AEO product, and the technical fixes — INP, schema, hreflang, llms.txt — that no other guide in the top ten actually publishes.

A few things you should know up front.

HubSpot's own customer data, published in the Spring 2026 Spotlight, shows organic traffic for HubSpot customers down 27% year over year.

Pew Research's July 2025 study of 68,879 real searches found Google CTR drops from 15% to 8% the moment an AI Overview appears — a 46.7% relative decline. Only 1% of users click links inside AI Overviews.

Zero-click search hit 69% by May 2025 (Similarweb).

And 42% of CRM software buyers now use AI search to evaluate vendors (HubSpot, January 2026 research).

HubSpot SEO is no longer one discipline. It's two — running in parallel, on the same site, with overlapping but distinct rules.

HubSpot SEO in 2026 is a "dual-track" discipline. You still need the technical fundamentals — Core Web Vitals (INP, not FID), schema, hreflang, internal linking — because 76.1% of URLs cited in AI Overviews also rank in the top ten organically.

Let's get into it. 

What does HubSpot SEO actually mean in 2026?

For ten years, "SEO" meant something simple: pick a focus keyword, write 800-1,500 words, get 2030 backlinks, watch the page climb to position three. Run the SEO Recommendations tool. Build a topic cluster. Done.

That playbook still works to some degree.

It just doesn't work as well as it used to — because the SERP itself changed.

25.8% of US searches now show an AI Overview (Similarweb, January 2026). Inside informational queries, that number is 39.4%. When an AI Overview shows up, classic click-through rate drops by roughly half. Ahrefs' December 2025 analysis of 300,000 keywords found the top-ranking page receives 58% less traffic when an AIO is present.

You're not losing rankings. You're losing the click.

The same shift hit at the platform level. HubSpot launched a dedicated AEO product in April 2026 — built on the XFunnel acquisition the previous October.

It tracks brand visibility across ChatGPT, Gemini, and Perplexity, ships with an AEO Grader free tool, and bundles into Marketing Hub Pro and Enterprise. HubSpot wouldn't have shipped a $50/month standalone AEO product if classic search were still doing all the work.

So the real question for HubSpot SEO now isn't "how do I rank?" — though you still need to.

It's "how do I get cited, mentioned, or chosen as the answer in the AI layer that's eating the SERP — while keeping the ranking work that still drives most of the traffic that converts?"

The next section is the part most agencies skip. 

Does traditional SEO still matter? Yes — and the overlap is the point

Every "AEO is the new SEO" think-piece you've read in the last twelve months is half right.

The half they get right: AI search is real, growing, and reshaping how people discover information.

The half they get wrong: that this means classic SEO is dead and you need a separate AEO playbook running in parallel.

Both halves miss the punchline.

Classic SEO still drives most of the traffic that converts. And almost every tactic that wins citations in AI answers also wins rankings in classic search. You're not running two playbooks. You're running one, with two output formats.

Three numbers tell the whole story.

1. AI engines pull from pages that already rank.

Growth Memo's analysis found 76.1% of URLs cited in AI Overviews also rank in the top 10 organically for the parent query. Surfer SEO's December 2025 study of 173,902 URLs adds the inverse: 68% of pages cited in AIOs were not in the top 10 for the parent query, but they were ranking — just for the related sub-queries the AI engine fanned out into.

Translation: if you're not ranking for anything, you're not getting cited for anything. Your AEO ceiling is bounded by your SEO floor.

The mechanism behind that overlap is retrieval-augmented generation — RAG. AI assistants like Google AI Mode, Perplexity, and ChatGPT search don't carry the entire web in their training data.

They run live retrieval against a search index every time they answer a query, then synthesise a response from what they pull back. The retrieval layer underneath is, with minor variations, the same problem you've been solving for years with classic SEO: crawlable site, indexed pages, fresh content, schema, internal links.

2. AI traffic is real but smaller than the noise suggests.

Google still controls roughly 89% of US web traffic (Digital Bloom, 2025) and processed over 5 trillion queries in 2024 — its first public disclosure since the 2-trillion figure in 2016.

SparkToro and Datos peg daily Google searches at 14 billion+, growing 21.6% year over year. Across all platforms, AI engines now account for roughly 6% of global search volume — small in absolute terms, but roughly triple what it was a year prior and accelerating.

ChatGPT has ~900M weekly users and 2.5B prompts a day, but only 34.5% of those prompts trigger a web search at all (Semrush, April 2026). Most ChatGPT use is a closed-loop conversation, not a click out. Google AI Mode and AI Overviews send more traffic — but their click-through rates are 1% and ~8% respectively (Pew Research, July 2025).

Classic Google organic still drives the lion's share of qualified click traffic for almost every B2B site we measure. AI search is real, growing, and worth investing in. It is not yet the dominant click source — and won't be for some time.

3. AI traffic converts much better — when you get it.

This is the part that flips the calculation. AI-referred traffic in the B2B portfolios we run converts at 4.4–5× the organic baseline — because the user has already had a conversational pre-qualification with an LLM before they click. They don't arrive at the top of the funnel. They arrive at "I think this is the tool, let me verify."

So the right way to think about it isn't AI vs classic. It's: classic SEO is the volume engine. AI search is the conversion-rate engine. You need both — and almost everything you do for one helps the other.

Here's the overlap, line by line:

Look at that table again.

Almost every row helps both columns.

Two practical implications fall out of that overlap.

The first: most "AEO" wins are unblocked SEO wins waiting to happen. Many HubSpot sites unknowingly block AI search crawlers in robots.txt — usually because someone copy-pasted a generic "block AI training bots" directive without separating training bots from retrieval bots. (We cover the three-tier breakdown — training vs search vs user — in section 9.) Fixing that overnight makes you eligible for citations you were already structurally qualified for, on content you'd already published.

The second: site speed is now an AEO ranking factor too. Mike King has documented AI engines hitting 499 client closed connection errors when a page is too slow to respond — the AI engine literally stops waiting and skips your content.

Slow LCP and INP don't just lose you Google rankings any more. They lose you AI citations the same week. Internal links carry double weight too — as Kevin Indig framed it, "internal links aren't just distributing authority; they're defining the semantic structure of your site." The semantic structure is what vector-search-based AI engines like Google AI Mode use to map your site into the entity graph they retrieve from.

The two genuinely AEO-specific tactics are at the bottom of the table: optimising paragraph length to the 40–60 word AI-extraction sweet spot (Princeton + iPullRank synthesis), and the llms.txt convention (which we treat as low-cost optionality — see section 10). Everything else is the SEO work you should already be doing, executed at higher quality.

That's why the framing of this guide is "dual-track" rather than "AEO replaces SEO."

You don't run two playbooks.

You run one — and you write to a higher standard, because the AI layer is more rewarding for citations, statistics, and structure than the classic SERP ever was.

Where the discipline genuinely diverges:

• The unit of optimisation shifts from page to passage. AI engines extract chunks. A 6,000-word pillar isn't a single ranking unit — it's twenty potential citation passages.
• The visibility metric shifts from rank to citation/mention. You can be cited without ranking position one. You can rank position one and never get cited.
• The off-page strategy shifts from backlinks to brand mentions. Reddit, YouTube, G2, Wikipedia, LinkedIn — that's where 89% of unbranded AI prompts get fulfilled.
• Volatility changes shape. Stable rankings give way to 40–60% citation drift month over month (AirOps research on 45,000 citations across 800 queries: only 1 in 5 brands stayed visible across 5 runs).

So yes — traditional SEO still matters. It still drives most of the click traffic, and it's the floor your AEO performance is built on. The honest answer to "should I focus on SEO or AEO?" is: focus on the content quality and technical foundations that win both, then add the four genuinely AEO-specific techniques (passage structure, inline citations as a primary tactic, off-site brand presence, and llms.txt) on top.

Content Hub's native SEO + AEO toolkit, feature by feature

HubSpot ships more SEO functionality than most teams use. Here's what's in the box, what each tool actually does, and where the limits sit.

The SEO tool. Path: Marketing > SEO. Page-level recommendations are on every plan. Topic clusters require Marketing Hub or Content Hub Professional or higher.

The crawler scans connected domains plus any external pages you add manually, and tags every recommendation with three dimensions: Impact (Low / Med / High), Technical Difficulty, and Role (Marketer / Developer). A full scan takes about three hours for a typical site, up to six for a large one.

The Optimize panel lives inside the page and blog editor. It surfaces page-specific recommendations, readability metrics for blog posts (reading level, ease, time, average word length, character count), and one-click "Fix in editor" actions for HubSpot-hosted content.

The Topics tool. Path: Marketing > SEO > Topic Clusters. Each topic equals one pillar plus up to 100 subtopic keywords. HubSpot validates that subtopic pages link back to the pillar — but only standard rich-text <a href> links count. CTAs and buttons are invisible to the validator. Search volume and Difficulty come from a Semrush integration, cached for 30 days.

Built-in analytics. Traffic Analytics, Sources, Pages, Topic Cluster reports, UTMs, Geography, Devices. Tracks Original Source vs Latest Source on every contact.

Attribution Reports. Contact-create attribution is on Marketing Hub Pro and Content Hub Pro. Deal-create and Revenue attribution require Marketing Hub Enterprise. Models include First-touch, Last-touch, Linear, U-shaped (40/40/20), W-shaped, Full-path, and Time-decay. Sampling caps at 100,000 associations per deal.

The Google Search Console integration. Marketplace install, OAuth connect, file-verified for HubSpot-hosted domains. Surfaces total pages indexed, impressions, clicks, average position, and per-page query data inside the Optimization tab on each page. The data purges 90 days after disconnect.

AEO Referrals. A traffic source bucket added to Reports > Marketing > Web Traffic Analysis > Sources during the 2025–2026 cycle. It separates AI-driven referrals (ChatGPT, Perplexity, Gemini search results) from generic direct traffic — partially. Most AI traffic still arrives without a referrer header and gets misclassified as Direct, which is the single biggest measurement caveat in this whole space.

Breeze AI runs across three layers in 2026. Breeze Assistant (renamed from Breeze Copilot) is the conversational sidekick — limits sit at 30 generations per minute and 1,000 per day. Breeze Agents are autonomous, installable from the Breeze Marketplace (public beta since INBOUND 2025). The Content Agent drafts blog posts, suggests topics from your top performers, auto-writes meta descriptions, and proposes internal links — Content Hub Pro/Enterprise plus credits. Breeze Intelligence handles enrichment, intent, and predictive scoring.

Brand Voice trains on a writing sample (HubSpot recommends 500+ words) and applies the tone to every AI-generated piece. Spring 2025 added the Brand Identity extension — auto-extracts tone, visual style, ICP, and competitive positioning. Spring 2026 launched Brand Voice trained on your published writing in private beta.

Content Remix repurposes a single asset (blog post, PDF, video, URL) into up to six outputs at once — social, ad copy, SMS, email, landing pages, podcasts. Maintains Brand Voice across all six.

AI Translation covers 60+ languages on the DeepL engine. Translates blogs, landing pages, website pages, plus author and tag translations. HubSpot auto-inserts hreflang in <head> for built-in multi-language variants. Important: hreflang cannot be added to sitemap.xml on HubSpot — confirmed limitation. Starter and Free are capped at three language variants per page.

HubSpot AEO. New product, launched April 14, 2026. Standalone at $50/month (no HubSpot subscription required), or bundled into Marketing Hub Pro (25 prompts/day) and Enterprise (50/day). Tracks ChatGPT (GPT-5.2), Gemini, and Perplexity. Claude is notably absent at launch. The free AEO Grader at hubspot.com/aeo-grader gives you a one-time score across five dimensions. 

Here's what's not in the box: keyword research at scale, daily rank tracking, backlink analysis, full technical site crawls, SERP-competitor content briefs, schema beyond BlogPosting and VideoObject, sitemap customisation beyond the four allowed tags, hreflang in sitemap.xml, per-page Core Web Vitals dashboards (the new Site Speed Dashboard from Spring 2025 helps but doesn't replace PageSpeed Insights), internal linking audit, and keyword cannibalisation reporting.

You'll still need Ahrefs or Semrush for the first four. Screaming Frog, Sitebulb, or Lumar for the technical crawl. Surfer, Clearscope, MarketMuse, or Frase for content briefs. Schema App or custom HubL for schema. PageSpeed Insights and the Calibre dashboard for ongoing CWV.

HubSpot SEO recommendations: how to use the tool without trusting it blindly

The SEO Recommendations tool is the workhorse most teams reach for first. It's useful. It's also wrong sometimes.

The tool checks roughly 25 things grouped under seven categories — Accessibility, Crawling and Indexing, Mobile Experience, On-Page SEO, Performance, Security, and User Experience. The full list is in HubSpot's KB article "Understand SEO recommendations" (last updated October 21, 2025).

Accessibility. Image alt text on every meaningful image, blank alt on decorative ones, lang attribute on <html>, no <meta http-equiv="refresh"> tag.

Crawling and indexing. Robots.txt blocking review, canonical link validation (it flags multi-canonicals, relative URLs, root-pointing canonicals, and cross-region canonicals), <a href> link presence (JavaScript-only links don't count), broken pages, hreflang validity, descriptive anchor text.

Mobile. Tap target ≥48×48 pixels, viewport meta tag.

On-page. Clear title, clear meta description, blog body word count ≥300 (counted inside the blog-post-wrapper class).

Performance. Correctly sized images and videos.

Security. All sub-resources load over HTTPS, rel="noopener" on every target="_blank" link.

Topic clusters and pillar pages, reframed for query fan-out

Topic clusters were made popular by HubSpot. The trigger was Google's Hummingbird (2013) and RankBrain (2015) updates. In 2026, the model still works — but it's doing a different job.

The original case for topic clusters was internal linking authority. Google sees you cover a topic comprehensively, treats you as the authority on that topic, ranks your pillar high.

That mechanic still works for classic SEO.

You're not optimising for one keyword anymore. You're optimising for the entire fan-out tree the AI engine generates from one parent query. Comprehensive subtopic coverage matters more than any single-keyword score.

Here's the workflow we run for clients on Content Hub:

1. Pick the head term you want visibility for.
2. Run the fan-out simulation. Use Qforia, or paste your head term into Gemini and Perplexity and capture the follow-up queries they suggest. You're looking for 30–60 sub-queries.
3. Cluster them. Group by intent. A cluster of "what is" questions becomes one section. A cluster of "how to" questions becomes another. A cluster of comparison questions becomes another.
4. Build the pillar around the clusters, not the keywords. This piece, for example, doesn't optimise for "HubSpot SEO" alone — it covers Content Hub features, AEO, INP, llms.txt, hreflang, schema, and measurement, because the fan-out for "HubSpot SEO" surfaces all of them.
5. Set up the cluster in HubSpot. Marketing > SEO > Topic Clusters > Add a topic. Attach the pillar URL. Add subtopic keywords for each cluster — capped at 100 per topic, with the validator only counting standard <a href> links from subtopic pages back to the pillar.

A practical note that HubSpot's own KB makes explicit but most agencies skip: "Creating topic clusters in HubSpot does not affect your website's SEO directly." The Topics tool is an organisational lens, not a ranking lever.

The lever is the actual content and the actual links. Performance Marketing Advisors recommends starting with 8–12 cluster articles per pillar instead of trying to fill all 100 slots — and we agree. 12 well-written, internally linked cluster posts will outperform 100 thin ones every time.

A few HubSpot-specific implementation rules:

• Build pillars on a custom website page template, not the default landing page template. Sticky tables of contents need CSS overrides the standard landing template doesn't allow.
• HubSpot has no native Table of Contents module. Use Magic ToC by Media Source, ToC by Helpful Hero, or Evoque Sticky Sidebar (~$20). HubL auto-ToC is hard because module IDs are auto-generated like content.widgets.module_163881811404819.body.main_title — fragile and ugly.
• Top-level placement, no form gating, no password protection. HubSpot's KB is explicit about this for pillars.
• Topic must appear in <title>, URL slug, and H1. Howells-Barby's old rule of thumb still applies: roughly one internal hyperlink per 150 words, with consistent anchor text across cluster posts.

Technical SEO on Content Hub: INP, hreflang, sitemap, jQuery

Technical SEO sounds intimidating. On HubSpot, most of it is already done for you.

What's left lives in three buckets: the platform defaults you can stop worrying about, the performance levers that actually move the needle, and the configuration gotchas that quietly break launches if nobody catches them.

Here's the version a RevOps lead can read without a developer in the room.

What HubSpot already takes care of

You don't have to think about most of the technical SEO checklist on HubSpot. The platform ships with:

• A global CDN with HTTP/2 — your pages load fast from anywhere in the world.
• Free SSL — the padlock in the browser, provisioned automatically when you connect a domain.
• Automatic image optimisation — uploaded JPGs and PNGs are served as WebP under the hood, sized for the device. Lighthouse may still flag "next-gen formats" because the URL still ends in .jpg. That's cosmetic, not real.
• Auto-generated sitemap at /sitemap.xml and editable robots.txt.
• Auto-emitted hreflang tags for every multi-language variant you create.
• Auto-applied BlogPosting schema on blog posts (and VideoObject when there's a video).
• Self-canonical tags by default on every page.

That's a strong baseline. Most of what's left is configuration — not engineering.

The performance fixes that actually move the needle

Google scores pages on Core Web Vitals — three numbers that measure how fast a page loads and how quickly it responds.

In plain English:

• LCP (Largest Contentful Paint) — how long before your biggest image or headline shows up. Target: under 2.5 seconds.
• INP (Interaction to Next Paint) — when a user clicks, taps, or types, how long before the page reacts. Target: under 200 milliseconds. INP replaced the older FID metric and is harder to pass — it measures the worst lag across the whole session, not just the first interaction.
• CLS (Cumulative Layout Shift) — how much things jump around as the page loads. Target: under 0.1.

The December 2025 Google core update made these scores visibly more important. Sites with slow LCP saw 23% more traffic loss. Sites with poor INP saw 31% drops.

On HubSpot specifically, three things drag those scores down more than anything else.

1. The HubSpot Chat widget. It's the single biggest INP killer we find on client sites. The fix is one developer ticket: delay the chat widget from loading until 3–5 seconds after the page is interactive. INP usually drops back into the green band the same week.

2. The jQuery default. HubSpot's default themes still load an old version of jQuery (a JavaScript library from 2011) on every page. It's slow and unmaintained. Settings > Website > Pages > jQuery. Uncheck the box. If a third-party script you can't remove still needs jQuery, your developer can load a modern version from the footer instead.

3. The default theme. The out-of-the-box HubSpot themes are built for design flexibility, not speed. For a pillar page or anything that has to rank, start from the CMS Theme Boilerplate — a free, performance-optimised theme HubSpot publishes on GitHub. One public case study saw a HubSpot site jump from a 36 to a 73 mobile Lighthouse score after rebuilding on the boilerplate. That kind of jump is the difference between Poor and Good on Google's scorecard.

The Spring 2025 Spotlight added a Site Speed Dashboard inside Content Hub that surfaces real-user CWV data. It's useful for spotting trends. For diagnosing a slow page, use Google's free PageSpeed Insights tool alongside it — both, not either.

The four configuration gotchas that silently break launches

These are the items that look fine in the UI but quietly cost you traffic. We catch them on every audit.

1. The sitemap doesn't include landing pages by default. HubSpot's sitemap auto-includes website pages and blog posts — but not landing pages. If you have campaign landing pages that should rank, add them manually inside the sitemap settings.

2. Robots.txt doesn't tell Google where the sitemap is. Most platforms add a Sitemap: line to robots.txt automatically. HubSpot doesn't. Add it yourself after launch under Settings > Content > Pages > SEO & Crawlers. We've seen launch checklists miss this for months.

3. Bulk redirects can silently fail. When you import a list of 301 redirects during a migration, there's a checkbox called "Disable redirect if a page exists at the redirected URL" that defaults to ON. If your destination URL exists at all — even as a draft — the redirect won't fire. Uncheck it before any bulk import, then re-enable case by case afterwards. This is one of the most common silent migration failures we see.

4. Cross-language links pollute your URLs. When you link from one language version to another, the link has to be marked with the target language code (hreflang="en", for example). Without it, HubSpot appends ?hsLang=en to the URL — which fragments the canonical signal and creates duplicate URLs in Google Search Console. Every link to a different-language variant should carry the matching language code, with hreflang="x-default" on your primary variant as fallback.

What HubSpot won't let you do (and what to do instead)

A few platform limits worth knowing about up front.

No custom sitemap.xml. You can edit which pages appear and how often, but you can't add image, video, news, or hreflang tags. The supported tags are basic. For most B2B sites this is fine.

No hreflang in sitemap.xml. HubSpot puts hreflang in the page <head> instead. That works for almost everything except very large multi-region sites with hundreds of language variants.

No multi-region pairs. HubSpot supports language variants (English, French, Spanish) but not regional pairs of the same language (en-US vs en-GB). If you have to differentiate by region, you'll need a workaround.

No translated system pages. 404, 500, search, and password-protected pages always render in English. If you serve a non-English market and care about this, a small JavaScript snippet can swap the copy based on the visitor's locale.

HubDB freshness limits. HubDB is HubSpot's database for templated pages — useful for location pages, service directories, or product listings. The catch: child pages inherit the parent's "last modified" date, not the row's. Search engines see the same freshness signal across hundreds of pages, which can hurt visibility. Worth knowing if you build a HubDB-driven site at scale.

That's the full technical SEO picture on HubSpot. Most of it is one-time configuration. Once it's right, you don't think about it again.

On-page SEO with the field limits HubSpot doesn't publish

HubSpot's own help docs are vague on field limits. Here's the actual list, pulled from the editor warnings and confirmed against the community ideas board.

A specific quirk worth flagging: by default, the blog H1 equals the page title. To set a different <title> for SERP display, use the now-live Customize page title option in the blog post Settings tab. This was a HubSpot Idea (262880) for years before it shipped.

Header tag management. Blog post title field becomes the H1 automatically. On website pages built with drag-and-drop, place a Heading module set to H1 at the top — never use two H1s. The SEO Recommendations tool flags missing H1s, multiple H1s, and skipped levels.

Internal linking. Only manual rich-text <a href> links count for the SEO Recommendations validator. CTAs, buttons, and Related Blog Posts modules don't count. The Related Blog Posts module is HubL-only — there's no drag-and-drop equivalent. Use this pattern for related-post grids

Image SEO. A late-2024 release made the File Manager Description field auto-populate as alt text site-wide (capped at 150 chars). Page-level alt overrides still work, but if you set the Description once in File Manager you don't need to re-enter it on every page. Huble flags any image over 100 KB in their audits — that's a reasonable threshold for B2B sites where most images are screenshots and headshots. 

AEO and GEO: the new layer, and what actually works

Three letters dominate the AI search conversation in 2026. They mean slightly different things.

SEO — ranking pages in classic SERPs. The discipline you already know.

AEO (Answer Engine Optimization) — structuring content so it gets extracted as a direct answer in AI Overviews, Perplexity, ChatGPT search, Gemini.

GEO (Generative Engine Optimization) — the academic term. Formalised by Aggarwal et al. (Princeton, Georgia Tech, IIT Delhi, Allen Institute) in arXiv:2311.09735, presented at ACM SIGKDD 2024. Optimising for citation, recommendation, or mention across generative AI.

There's also LLMO (LLM Optimization) and Mike King's umbrella term Relevance Engineering (iPullRank, 2025), which he defines through "vector embeddings, cosine similarity, topic segmentation, and semantic scoring."

Pick whichever label your team prefers. The tactics underneath are the same.

The AI search ecosystem in April 2026. ChatGPT runs at ~900M weekly active users (Feb 2026), 2.5B prompts a day. Only 34.5% of those queries trigger a web search (Semrush, April 2026 — down from 46% in late 2024). Market share dropped from 87.2% in January 2025 to 68% in January 2026 (Similarweb) as Gemini ate share.

What actually works to win citations:

• Cite sources (inline citations to authoritative third parties): +115% visibility for lower-ranked pages.
• Add statistics (specific data points and percentages): +41% visibility.
• Add quotations (attributed expert quotes): +28–37%.
• Fluency optimisation and easy-to-read language: incremental gains.

Other research-backed tactics:

• Brand search volume has a 0.334 correlation with LLM citations — outweighing traditional backlinks (Growth Memo, SE Ranking).
• High-traffic domains earn 3× more AI citations (SE Ranking analysis of 2.3M pages, SHAP value 0.63).
• 65% of AI bot hits target content from the past year — recency matters more for AI than for classic search.
• Listicles dominate citations: 25% citation rate generally, 32.5% in some studies, 40.86% on commercial queries. Articles win informational queries at 45.48%.
• Optimal paragraph length for AI extraction is 40–60 words (Princeton + iPullRank synthesis). Longer paragraphs get chunked or skipped.
• A brand mentioned on 4+ external platforms is 2.8× more likely to be cited in ChatGPT (Digital Bloom).
• 89% of unbranded prompts are fulfilled by third-party sources (Bain/ScrunchAI). For category-level prompts, owned content is just 4.3% of citations — but those URLs land in the top 5% of all cited domains.

Reddit, LinkedIn, YouTube, Wikipedia, G2, and Forbes top the citation lists across LLMs. Wikipedia is the #1 source on ChatGPT (7.8% of citations). Perplexity cites social at 19.4% vs ChatGPT's 5.3%. ChatGPT leads institutional citations at 15.3%, mostly Wikipedia.

The robots.txt decision — three-tier system. AI bots split into training, search, and user-triggered roles. You can allow retrieval bots while blocking training bots. The canonical pattern when you want citation visibility without contributing to model training:

ClaudeBot blocking rates among top news sites hit 69% (BuzzStream, January 2026). GPTBot blocking sits at 62%. OAI-SearchBot at 49%. The training-vs-retrieval split lets you hold a stronger position without losing citation access.

Conversational query reality. AI search queries average 70–80 words vs 3–4 for classic searches — a 17–26× increase in query complexity (iPullRank, December 2025). And 65–85% of ChatGPT prompts have no matching keyword in Semrush's database (Semrush, April 2026). Keyword tools alone now miss the majority of AI search demand.

Where do you find the actual question phrasing? Mine GSC's queries report. Mine People Also Ask boxes. Mine your customer support tickets and sales call transcripts. Those are the real questions. That's the language to write into your subheaders, FAQ blocks, and bolded Q-format pivots.

llms.txt on HubSpot: the workaround

llms.txt is the Markdown file at the root of a domain that lists curated links to important pages — a kind of robots.txt for LLMs. Jeremy Howard (Answer.AI) proposed it on September 3, 2024. The companion llms-full.txt file holds the full documentation in flat Markdown.

By April 2026, BuiltWith's October 2025 scan found 844,000+ websites publishing llms.txt. SE Ranking saw 10.13% adoption across 300,000 domains analysed. Anthropic, Cloudflare, Stripe, Mintlify, Zapier, Vercel, Bolt.new, and Windsurf all publish one. Yoast added auto-generation. Webflow added root upload support.

Now the honest part.

John Mueller (Google): "AFAIK, none of the AI services have said they're using LLMs.TXT (and you can tell when you look at your server logs that they don't even check for it)."

Ahrefs: "No major LLM provider currently supports llms.txt. Not OpenAI. Not Anthropic. Not Google."

Search Engine Land tested 9 sites and 8 saw no measurable traffic change after implementation. Google has been explicit that llms.txt is a proposal and not a standard, comparing it to the discredited keywords meta tag.

The pro side: Profound claims OpenAI and Microsoft bots fetch llms.txt files. Anthropic's Alex Albert publicly endorsed it. Vercel claims 10% of new signups now arrive from ChatGPT.

Our verdict: implement it as low-cost optionality, not a primary lever. It takes thirty minutes. The downside is zero. The upside — if any major LLM provider formalises support — is small but real.

Here's the HubSpot-specific catch.

HubSpot does not natively support root-level non-HTML files. Confirmed by HubSpot's own product team in community thread 1136969:

"We've done a few deep dives on our web traffic data and as of now, we're not seeing enough of an indication that answer engines are using llms.txt as a convention for crawling."

That's HubSpot's official line. They're not building it.

The workaround we publish for clients — and have not seen anywhere else in the top twenty HubSpot SEO guides — is a four-step File Manager + URL Redirect pattern:

1. Generate llms.txt locally. Use a free generator or a quick LLM prompt. Keep it short — page title, URL, one-line description for each priority page. For docs-heavy sites, generate the long-form llms-full.txt too.
2. Upload to HubSpot File Manager. Files > Files & Templates > Files. Note the public URL — it'll be on f.hubspotusercontent10.net or your custom CDN domain.
3. Create a 301 URL Redirect. Settings > Content > Domains & URLs > URL Redirects. Source: https://yourdomain.com/llms.txt. Destination: the File Manager URL.
4. Test with curl. Run curl -I https://yourdomain.com/llms.txt — you should see a 301 followed by a 200 returning the file content.

For ongoing automation, use the HubSpot Files API v3 to overwrite the file when you publish new content. We've wrapped this in a small workflow that runs nightly: pull the latest published pages from HubSpot, regenerate llms.txt, upload via API, done.

That's the full workaround. Cheap. Safe. Optional. And the only HubSpot-native llms.txt deployment guide we've seen published in 2026. 

The new HubSpot AEO product, in plain English

HubSpot launched a dedicated AEO product on April 14, 2026 — built on the XFunnel acquisition (October 31, 2025; Israeli AEO startup founded January 2025 by Beeri Amiel and Neri Bluman).

Yamini Rangan's framing on the launch:

"How buyers search is fundamentally changing. They are asking questions in places like ChatGPT and Gemini, and the companies that show up in those answers are already winning."

Here's what the product actually does.

Pricing. $50/month standalone (no HubSpot subscription required). Included in Marketing Hub Professional with 25 prompts/day. Included in Marketing Hub Enterprise with 50 prompts/day. Additional answers are $20/mo per 1,000.

Tracked engines. ChatGPT (GPT-5.2), Gemini, Perplexity. Claude is not currently tracked — flag this if Claude is a meaningful share of your buyer journey.

Metrics. Brand Visibility %, Share of Voice, Citation Influence Rate (HubSpot's proprietary metric — how often AI-cited sources reference your brand), Citation analysis by content and source type, Sentiment with theme tagging, Recommended Actions tied to Breeze Assistant content drafts, and CRM-data-aware prompt suggestions on Pro and Enterprise.

The Citation Influence Rate metric is the one worth understanding. Most AEO tools measure whether your domain shows up in AI answers. CIR measures something different: when an AI engine cites a third-party source (a Reddit thread, a G2 review, a Forbes article), does that source mention your brand? It's a proxy for the off-site brand presence that the Princeton GEO research showed actually moves citation rates.

Free AEO Grader. hubspot.com/aeo-grader scores your brand once across five dimensions: sentiment, presence quality, brand recognition, share of voice, market position. Includes a 28-day free trial of the full product with 10 ChatGPT prompts. Run yours before you decide whether to buy the paid product.

HubSpot's own claim — and treat this as marketing data: AEO beta customers saw 20% more traffic from AI vs non-users. HubSpot's internal AEO strategy delivered a 1,850% increase in qualified leads. The 1,850% number is HubSpot-claimed, not third-party verified, and reflects a specific baseline. Useful directionally. Don't quote it as a benchmark.

Should you buy it? A useful decision rule:

• Yes, if you're already on Marketing Hub Pro or Enterprise — it's bundled, the only cost is your team's attention.
• Yes, if you're a HubSpot-native shop without a third-party AI tracking tool, and $50/month is in budget.
• Maybe, if you're already paying for Profound, Peec, AthenaHQ, or Otterly.ai — HubSpot AEO covers a narrower set of engines than Profound's 400M+ prompt database, so you may want both for now.
• No, if Claude visibility is critical to your audience and HubSpot hasn't added it yet (check the latest changelog).

That's the product. Use it as one input — not the only one — into your AI visibility programme.

Measurement: KPIs, GSC, AI-citation tracking

If you can't measure it, you can't defend the budget. The 2026 HubSpot SEO measurement stack has eight KPIs worth tracking.

1. Organic sessions. The baseline. Contextualise against the -27% YoY HubSpot benchmark — if you're flat, you're outperforming. If you're up, you're winning.
2. Organic-attributed revenue / pipeline. From HubSpot Attribution Reports. Revenue model requires Marketing Hub Enterprise. Sampling caps at 100K activities per deal — a real constraint on enterprise reporting.
3. AI Referral sessions and conversion rate. Reports > Marketing > Web Traffic Analysis > Sources > AI Referrals. AI traffic converts at 4.4–5× the organic baseline in the B2B portfolios we've measured. That ratio is the single most under-appreciated number in marketing analytics in 2026.
4. Branded vs non-branded organic. Branded growth is the strongest predictor of LLM citations (0.334 correlation, outweighing backlinks). Track branded query volume in GSC monthly.
5. AI citation share / mention rate / sentiment. Via HubSpot AEO, Profound, Peec, AthenaHQ, or Otterly.ai. Pick one — running multiple is useful but expensive.
6. Citation Influence Rate. HubSpot's proprietary metric. Tracks third-party brand mentions inside AI-cited sources.
7. Share of voice in AI answers. Vs your tracked competitors.
8. Crawl-to-referral ratio by bot. Cloudflare baselines: ClaudeBot at 20,583:1, GPTBot at ~1,700:1. If your numbers are wildly worse, you're being crawled without being cited — which usually means you're not on the platforms LLMs cite (Reddit, YouTube, Wikipedia, G2).

The single biggest measurement caveat. 70.6% of AI traffic arrives without a referrer header and gets misclassified as Direct in GA4 (Digital Bloom, February 2026). Standard analytics dramatically underestimate AI-driven traffic. Until GA4 adds proper AI source detection, treat your "Direct" channel as 50–70% AI traffic for any content-heavy B2B site.

The reporting stack we run for clients:

• HubSpot dashboards for sessions, attribution, topic-cluster performance, AEO Referrals.
• Google Search Console for queries, impressions, CTR, average position, indexing health.
• HubSpot AEO for AI visibility on ChatGPT, Gemini, Perplexity.
• A third-party AI tracker (usually Profound or Peec) for cross-platform depth — including Claude.
• Custom HubSpot properties to flag "AEO-optimised" content vs legacy, so you can track lift over time.

That's enough. Don't add more tools.

What to stop doing in your old HubSpot SEO playbook

Six things were defensible in 2022, are wrong in 2026, and still appear in agency guides ranking for "HubSpot SEO" today.

Stop optimising for FID. Interaction to Next Paint replaced First Input Delay as a Core Web Vital. INP measures the worst latency across the whole session — it's harder to pass, and it's the one Google now scores. Any guide still listing FID is using a metric that no longer exists.

Stop turning AMP on for new sites. AMP has been deprecated as a ranking signal for years. On HubSpot specifically, AMP strips tracking JS, breaks GTM, disables CTAs and forms. Turn it OFF for new sites. If you've got it on legacy blog posts, leave it for now and remove it the next time you redesign the template.

Stop quoting the "1,890 words is the first-page average" stat. That was a 2016 SerpIQ study. The 2026 reality is messier — informational pillars trend long (3,000+) while listicles and answer-format pages can win at 800. Match the format to the SERP.

Stop building topic clusters to fill 100 subtopic slots. HubSpot's tool allows it. Performance Marketing Advisors recommends 8–12 cluster posts per pillar, and we agree. Twelve well-written, internally linked posts beat 100 thin ones every time.

Stop trusting the SEO Recommendations tool as a source of truth on indexing. Cross-reference every "noindex" or canonical alert with GSC's URL Inspection before you act on it. False positives are documented and reproducible. 

The Superwork HubSpot SEO Maturity Model

Most HubSpot sites we audit sit at Level 1 or 2. Here's the five-level model we use to score where a portal is and what to fix next.

Level 0 — Default. Out-of-the-box theme. SEO Recommendations never run. No GSC integration. Blog posts auto-publish without meta descriptions. Topic clusters tool unused. What to do: install GSC, run the first SEO scan, and write meta descriptions on the top 20 pages by traffic.

Level 1 — Optimised. SEO Recommendations run weekly. Page-level meta and headings done correctly. Image alt text in File Manager. Schema = BlogPosting auto-toggle on. Topic clusters set up but not maintained. What to do: audit INP and CWV, fix the jQuery default, layer Organization and FAQPage schema.

Level 2 — Tracked. GSC integrated. HubSpot Analytics dashboards built for organic performance. Topic clusters maintained quarterly. Pillar pages exist. Internal linking systematic. What to do: stand up Attribution Reports (requires Marketing Hub Pro for contact-create, Enterprise for revenue), connect organic traffic to deal stages.

Level 3 — Attributed. Attribution Reports running on contact-create and revenue. Marketing can show CFO the channel-to-deal map. Branded vs non-branded tracked. Topic-cluster ROI measured by cluster, not by post. What to do: layer AEO. Install HubSpot AEO or third-party. Audit off-site brand presence. Write to the Princeton GEO findings (citations, statistics, quotes).

Level 4 — AI-Visible. AEO citations tracked monthly. llms.txt deployed via the File Manager workaround. Schema layered (FAQPage, HowTo, VideoObject, Product, Service). Off-site presence on Reddit, G2, YouTube, Wikipedia. Breeze Agents in the content workflow. CWV in the green. INP <200ms on mobile. AEO Referrals separated from Direct in GA4 attribution model.

A typical client engagement starts at Level 1 or 2. The 30/60/90 path to Level 3:

• Days 1–30: SEO scan + GSC integration + meta sweep on top 20 pages + INP audit + schema layering on the homepage and pillar templates.
• Days 31–60: Attribution Reports stood up + topic cluster restructuring around fan-out queries + llms.txt workaround deployed + first AEO Grader run.
• Days 61–90: First branded vs non-branded report + AI Referrals dashboard + Citation Influence Rate baseline + 90-day review and roadmap to Level 4.

That's the framework we run. It's deliberately less ambitious than what HubSpot AEO marketing suggests — because most teams can't sustain Level 4 work without first doing Levels 0–2 properly. 

Where this approach doesn't fit

We're explicit about the cases where the playbook in this guide is the wrong call.

You're not on HubSpot and don't intend to migrate. The Content Hub-specific tactics — the SEO Recommendations tool, the Topics dashboard, Breeze Agents, the Files API workaround for llms.txt — don't translate to WordPress, Webflow, or a custom stack. The AEO and GEO sections still apply universally; the rest doesn't.

You're a very early-stage startup with no marketing function. This guide assumes you can dedicate someone to running a measurement and content cadence. If you're three founders shipping product, your highest-impact SEO move is probably a single excellent comparison page and a Reddit/X presence — not a 12-cluster pillar architecture.

You're a pure e-commerce operator. Most of the AEO research findings still apply, but commercial-intent queries behave differently in AI Overviews (4% appearance rate vs 39.4% for informational). Product schema, review schema, and merchant feeds matter more than what we've covered here.

You sell into a market where Claude is the dominant AI assistant. HubSpot AEO doesn't track Claude. You'll need a third-party tracker (Profound, Peec, AthenaHQ) regardless. Don't buy HubSpot AEO as your only measurement layer in this case.

Your CMO's KPI is still "rankings." This playbook will frustrate you. Rankings still matter, but they're a leading indicator of citation share, not the outcome itself. If your reporting can't shift, the dual-track model will look like extra work for unclear payoff. The reframe has to come first.

Frequently Asked Questions

Is HubSpot good for SEO in 2026? Yes — for content-driven B2B and B2C sites. Content Hub ships strong defaults: SSL, CDN, auto-WebP, automatic hreflang, BlogPosting and VideoObject schema, the SEO Recommendations tool, and Topic Clusters. The limits are sitemap customisation, schema beyond two types, and rank tracking — all of which require third-party tools to fill. For B2B mid-market companies already running HubSpot CRM, it's the right call.

Does HubSpot do SEO automatically? Partially. HubSpot handles SSL, sitemap generation, image WebP conversion, BlogPosting schema (post-August 2022 blogs), VideoObject schema (since May 2025), hreflang in <head> for multi-language variants, and a CDN with HTTP/2. It does not write meta descriptions, generate FAQPage schema, optimise INP, fix the jQuery default, or get you cited in AI Overviews. The defaults are good. The strategy is yours.

What is HubSpot AEO and is it worth $50/month? HubSpot AEO is the dedicated AI search visibility product launched April 14, 2026, built on the XFunnel acquisition. It tracks brand citations, share of voice, and sentiment across ChatGPT, Gemini, and Perplexity (not Claude). $50/month standalone, included in Marketing Hub Pro and Enterprise. Worth it if you're already on Pro or Enterprise (it's bundled), if you don't have a third-party tool, and if Claude isn't critical to your audience. Run the free AEO Grader at hubspot.com/aeo-grader before you decide.

Can I add llms.txt to HubSpot? Not natively. HubSpot doesn't support root-level non-HTML files. The workaround: generate llms.txt locally, upload to File Manager, create a 301 URL Redirect from yourdomain.com/llms.txt to the File Manager URL, then test with curl -I. Automate ongoing updates via the HubSpot Files API v3. Treat it as low-cost optionality — most major LLM providers haven't formally adopted the standard yet, but the implementation cost is roughly thirty minutes.

Does HubSpot support hreflang? Yes, automatically — for multi-language variants created through HubSpot's multi-language groups. Hreflang is auto-inserted in <head> using full URLs. Multi-region pairs (en-US vs en-GB on the same content) aren't well supported. Hreflang cannot be added to sitemap.xml. For internal links across language variants, set hreflang="en" (or your target language) on the <a> tag to prevent HubSpot appending ?hsLang= query parameters that fragment your canonical signal.

What's the right pillar page length for HubSpot SEO in 2026? Match the format the SERP rewards. Informational pillars covering broad topics (like this one) tend to land between 4,000–6,000 words. Commercial-intent comparison pages can win at 1,500–2,500. Listicles dominate AI citations at 25% and can be much shorter. Drop the "1,890-word average" rule — it's a 2016 stat that doesn't survive contact with the 2026 SERP.

Do topic clusters still work in 2026? Yes, but the job changed. The original case was internal linking authority for classic SEO. The 2026 case is comprehensive subtopic coverage to win Google's AI Mode query fan-out — which fires up to 16 follow-up queries per parent question. Build clusters around the fan-out, not around 100 long-tail keywords. 8–12 well-written, internally-linked cluster posts per pillar beats 100 thin ones every time.

The bottom line

HubSpot SEO in 2026 is a dual-track discipline.

You still need the technical fundamentals — Core Web Vitals (INP, not FID), hreflang, schema, internal linking, the SEO Recommendations tool used as a triage feed and not a source of truth. Those still drive 76.1% of AI Overview citations. Skip them and you exit both layers of the SERP.

But you have to layer the AI track on top. Passage-level structure. Inline citations. Statistics and attributed quotes (the Princeton GEO findings). Off-site brand presence on Reddit, YouTube, G2, and Wikipedia. The llms.txt workaround. Schema beyond BlogPosting. The new HubSpot AEO product, used as one signal among several. AI Referrals tracked separately from Direct in your attribution model.

That's the whole brief. Most agencies covering "HubSpot SEO" still write half of it — usually the half that was true in 2022.

If you want a one-page version of every check we run on a HubSpot portal — INP, hreflang, schema, llms.txt, the AEO citation baseline, all of it — download the free HubSpot SEO + AEO Checklist. It's the same audit sheet we use on day one of every Superwork engagement.

If you'd rather have us run it on your portal directly, talk to Superwork about a HubSpot SEO audit. We're built to run the dual-track playbook this guide describes — classic search foundations plus the AEO layer that's eating the SERP.

That's HubSpot SEO in 2026.