Use Semrush AI Visibility and HubSpot Influence to Pick Revenue-First Content Refreshes

Most AI visibility reports die in a dashboard.

Someone exports a Semrush chart, circles that the brand appears in 14% of ChatGPT answers, and then the team refreshes whatever page feels easiest. I get why that happens. Semrush made the AI search layer visible in 2025, and the screenshots look board-ready. But visibility by itself is a weak operating metric. The better move is to join AI mentions to HubSpot campaign influence, then refresh the pages already touching pipeline.

That is where the money is.

Semrush launched its AI Toolkit on March 4, 2025 to track how systems like ChatGPT, Gemini, and Perplexity describe brands. Its current AI Visibility material now includes ChatGPT, Gemini, Google AI Mode, Google AI Overviews, and Perplexity across reports like Visibility Overview, Brand Performance, Competitor Research, Prompt Research, and Prompt Tracking. Google matters here because AI Mode rolled out in the U.S. on May 20, 2025, with follow-up questions, links to the web, and query fan-out under the hood. That changes the content job. A page is no longer fighting for one blue link. It may become one cited source inside a longer answer.

HubSpot brings the missing second half. Its campaign tool can show influenced contacts for contacts that engaged with campaign assets, and HubSpot lists this capability for Marketing Hub Professional and Enterprise accounts in its August 3, 2026 influenced contacts documentation. For operators with the right portal setup, campaign reporting can also connect campaigns to contacts, deals, and revenue through custom reports or attribution reports. The campaign metrics API even exposes fields like influencedContacts, sessions, newContactsFirstTouch, and newContactsLastTouch.

So the workflow is not “find AI gaps, write content.” It is “find AI gaps around pages that already create or influence revenue.” That distinction saves months.

Start With Pages, Not Prompts

I would not begin inside Prompt Research. I would begin with a URL export.

Pull the cited pages and mentioned topics from Semrush AI Visibility Toolkit. The free checker says it can show total brand mentions, mention breakdowns across ChatGPT, Gemini, and Google results, topics and prompts where the brand appears, and pages that drive AI citations. The paid AI Visibility Toolkit goes deeper. As of Semrush’s current pricing page, the Base AI Visibility plan is listed at $99 per month per domain when billed annually, with 25 custom prompts for daily AI rankings and mentions from ChatGPT, Google AI, Gemini, and Perplexity.

That gives you a workable dataset. For each URL, capture the page path, AI platform, prompt or topic, citation count, mention count, sentiment if available, and the competitor that appears when you do not. Keep the export ugly. A Google Sheet with 11 columns is enough for the first pass.

Then pull HubSpot campaign influence by page group. In a clean HubSpot portal, the better structure is one campaign per content cluster, not one campaign per asset. For example, a B2B payroll startup might have campaigns named “Payroll Compliance Hub Q3 2026,” “Contractor Payments Hub Q3 2026,” and “Benefits Admin Hub Q3 2026.” Each campaign owns blog posts, landing pages, ads, forms, CTAs, and nurture emails tied to that theme.

Now you can ask a sharper question. Which pages are visible or nearly visible in AI answers and tied to influenced pipeline?

The Join That Changes The Queue

Here is the join I use in client work.

First, map each Semrush URL to a HubSpot campaign. If one page belongs to multiple campaigns, pick the campaign that owns the buyer intent. A comparison page like /compare/rippling-vs-gusto/ belongs with the competitor campaign, even if three email newsletters linked to it.

Second, export influenced contacts, influenced deals, and influenced revenue from HubSpot for the last 180 days. I like 180 days because 30 days is too twitchy for B2B, and 365 days lets stale campaigns hog the queue. For ecommerce or PLG with fast conversion, use 60 or 90 days.

Third, add Search Console clicks and conversions if you have them. This is not because Search Console can see ChatGPT. It cannot. It gives you the organic baseline. If a page has 2,400 Google clicks, 19 HubSpot influenced deals, and 0 citations in AI Mode for its core prompt, that page is a refresh candidate. If another page has 43 AI mentions and $0 influenced pipeline, it goes lower unless the founder needs brand coverage for a funding narrative.

In one example account, a cybersecurity SaaS company had a “SOC 2 automation checklist” page with 1,860 organic clicks in July 2026, 74 influenced contacts in HubSpot, 11 influenced opportunities, and $312,000 in influenced pipeline. Semrush showed the brand appearing in ChatGPT for branded SOC 2 prompts, but missing from generic prompts like “best SOC 2 automation tools for startups” where Vanta and Drata appeared. That page went to the top of the refresh queue.

A glossary page on “audit evidence” had 31 AI mentions and only two influenced contacts. It looked good in the AI dashboard. It did not earn a sprint slot.

Score Refreshes Like An Operator

I use a 100-point score. It is plain enough that a founder can argue with it in five minutes.

AI opportunity gets 30 points. Give more weight to prompts where competitors are cited and you are absent, where your page is already cited but the answer describes you poorly, or where Google AI Mode shows competitor links for a category you sell into.

Revenue influence gets 40 points. HubSpot influenced pipeline should carry the most weight because the content team is not paid in mention counts. A page tied to $100,000 in influenced pipeline should beat a page tied to $4,000 unless the smaller page has strategic value, such as a new product launch.

Refresh feasibility gets 20 points. Some pages need two paragraphs, new screenshots, and schema cleanup. Others need original data, customer quotes, legal review, and a product marketer who is booked until October. Score the work honestly.

Measurement clarity gets 10 points. If the campaign, URL, prompt set, and HubSpot influence window are clean, you can learn from the refresh. If attribution is smeared across six campaigns and three cloned landing pages, fix tracking before rewriting.

The output is a ranked queue with numbers attached. A page scoring 82 gets refreshed before a page scoring 51. Nobody has to perform strategy theater in a content meeting.

What To Change On The Page

AI answer systems need extractable evidence. That does not mean stuffing a page with FAQ blocks until it reads like a help center after a long weekend. It means the page should answer the prompt directly, name the category, show proof, and make the comparison easy to parse.

For a HubSpot campaign page about “B2B demand generation benchmarks,” I would add a 2026 benchmark table with sample size, company segment, channel, spend band, conversion rate, and source methodology. I would update the page title to match the actual buyer question. I would add a short section comparing paid search, LinkedIn Ads, partner webinars, and organic search for companies between $2 million and $20 million in ARR. I would cite first-party data if the company has it. If not, I would use named third-party sources like HubSpot, Semrush, Google, LinkedIn, or Gartner and link to the exact report.

Freshness matters too. Semrush’s AI Visibility guidance flags AI readiness issues like outdated last-modified headers, robots.txt blocks, missing llms.txt files, and pages that are too long for AI models. I would not treat llms.txt as magic. I would treat it as a crawl and clarity task, sitting next to sitemap hygiene, canonical tags, server-rendered content, and schema that matches the page.

The page still needs to sell. Add proof where the reader is deciding. That might be a HubSpot form tied to the same campaign, a demo CTA, a customer example, a pricing caveat, or a product screenshot from September 2026 instead of a UI shot from 2023. AI visibility work that removes conversion paths is just SEO cosplay with a new costume.

Build The Reporting Loop

After the refresh, track three dates.

Record the publish date in the CMS, the campaign asset date in HubSpot, and the prompt tracking start date in Semrush. If the page was refreshed on September 6, 2026, do not judge it on September 9. I use a 28-day check for crawl and citation movement, then a 90-day check for influenced contacts and pipeline. B2B sales cycles are slow. The dashboard should admit that.

In HubSpot, keep the refreshed page attached to the same campaign unless the positioning changed enough to justify a new campaign. Splitting the asset too early makes the before-and-after read muddy. In Semrush, pin 10 to 25 prompts that match buyer language, not internal category names. “Best customer onboarding software for Series A SaaS” beats “customer lifecycle enablement platform” unless buyers actually say that.

The reporting view I want is small. Page URL. Target prompt. AI platform. Current mention status. Cited page status. HubSpot influenced contacts. HubSpot influenced pipeline. Refresh owner. Publish date. Next check date.

That is the dashboard. Ten columns, reviewed every other Tuesday.

The Mistake To Avoid

The easiest mistake is chasing prompts with no commercial spine.

A founder sees that a competitor appears in ChatGPT for “what is revenue marketing” and wants a 3,000-word refresh by Friday. Maybe that is worth doing. But if the existing page has no HubSpot campaign, no form assists, no influenced deals, and no clear buyer path, the first job is not the rewrite. The first job is wiring the page into the revenue system so future work can be judged.

This is where marketing operators beat content calendars. They know the CRM is messy. They know paid and organic overlap. They know one page can influence a LinkedIn retargeting click, a demo form, and a sales email three weeks later. Semrush shows where AI systems are narrating the market. HubSpot shows whether that market narrative is near revenue.

Prioritize the overlap. Refresh there first.