AI SignupsPlaybook

I’m Brian Shelton. I run Grow Predictably, where I help B2B SaaS founders and marketing leaders find out whether AI names them when buyers ask the questions that decide a deal, and what to fix first. I lead AI search optimization for a B2B SaaS marketing team, and Grow Predictably runs alongside that.

I put this together because the most useful evidence in AI search right now isn’t a score from a tool. It’s a handful of companies who measured their own signups and then wrote down what they did.

Want to know where you show up today? Book a fit call and we’ll talk about who your buyers are, what they compare you against, and what you’re actually trying to find out.

Four well known SaaS companies that publicly shared how many of their signups or new users come from AI, and what each one says it did about it.

Every number here comes from the company itself: a CEO’s post, a company blog, a founder’s milestone update. Every one is dated and linked in the Sources section at the bottom. Where a company never explained how it happened, this says so instead of guessing.

None of these companies are my clients. And nobody, including them, can promise what an AI tool will say about you. This is what they’ve shared themselves.

These numbers were measured four different ways. Vercel and Ahrefs counted signups by where the visitor came from. Webflow reported a share of total new signups. Tally asks every new user where they found it. That means the numbers are worth reading side by side, but not worth ranking against each other.

These numbers are from different dates. The oldest is from March 2025 and the newest is from September 2026. AI search moves fast. Treat each one as true on the date it was published.

Referral tracking undercounts. Not every AI tool passes referral data to your analytics, and plenty of buyers read an answer, then type your name into Google later. Whatever your analytics shows for AI is probably a floor, not the full number.

Plan about 45 minutes. About 25 to read, and 20 to go through the questions at the end with whoever owns AI search on your team.

1. Read the at-a-glance table first. It gives you the four numbers, the dates and how each was measured in one place
2. Read the four company sections in order. Each one has the number, how it was measured, what the company says it did, and what’s worth taking from it
3. Read “What all four have in common.” Only patterns at least three of the four companies stated themselves made the list
4. Take the questions to your team. That’s the part that turns this from reading into a plan
5. Measure your own number this month. The last section shows the two cheapest ways to start

⚠️ Don’t skip to the patterns. The patterns only hold up because of the sourcing in each section. If you lift them into a leadership deck without the numbers and dates behind them, the first sharp question will take them apart.

The number. In March 2025, Vercel CEO Guillermo Rauch posted that ChatGPT referred 4.8% of new Vercel signups, and that it had been under 1% six months earlier. A month later, on April 9, 2025, he posted that it had reached 10%.

How it was measured. Signups referred by ChatGPT. This counts ChatGPT only, so other AI tools would add to it.

What Vercel says it did. In June 2025, Vercel published its own write-up on adapting SEO for AI search. The main points:

• It isn’t a replacement for SEO. Vercel treats it as an adaptation. Crawlable, indexable pages, a clean heading structure and fresh content still matter to both
• Pick a concept you can own. Vercel’s view is that models tend to favor the first or the clearest explanation of a topic. So find questions where competitors are thin or absent, ideally ones your users already ask in places like Reddit, GitHub, Discord and X
• Publish the definitive version. Go deep, include original data, code, tables or customer stories, and use the same terms consistently. Their test: could a competitor copy this tomorrow? If yes, go deeper
• Make it readable by machines. Vercel notes that most AI crawlers fetch pages without running JavaScript, so content should be served as plain HTML through server-side or static rendering. Confirm the pages are indexed in both Bing and Google
• Earn real mentions, not paid ones. Share in community threads, changelogs and demos, with Reddit, GitHub, Hacker News, X, LinkedIn and Stack Overflow named as high-signal places. Vercel advises against paid links
• Refresh on a schedule. Review content at 30, 90 and 180 days, fix broken pages, and archive what’s outdated

Worth taking from it. Vercel’s advice comes down to answering the specific questions your buyers ask, in more depth and more clearly than anyone else, in a form machines can read. Your buyers ask specific questions too. The question is whether your site has the clearest answer to them.

Worth knowing. Vercel says plainly that nobody has all the answers yet and that there’s no reliable dashboard for this. Its tracking advice is to watch AI referrers, check citations by hand, watch community mentions and keep an eye on index coverage.

The number. Webflow’s AEO/SEO lead, Vivian Hoang, wrote that as of June 2025, 8% of Webflow’s total new signups came from AI, compared with 2% in October 2024.

How it was measured. Share of total new signups from AI. The post doesn’t spell out the attribution method.

One thing to note. The post’s subtitle describes the growth as 4x in six months, while the dates in the body span about eight. Either way, the growth happened in under a year. This playbook uses the dates in the body.

What Webflow says it did. The post describes two goals.

Goal one, more brand visibility. Webflow’s marketing team works to get the brand mentioned in the sources AI tools lean on: tech publications, Reddit, YouTube and industry sites. Webflow describes this as a whole-team job, not just an SEO one.

Goal two, better owned content. Since Webflow can’t see the prompts people type, it studies the pages that already bring in AI traffic and conversions, and works backward from those. The practices it lists:

• Meaningful content without filler
• Answering the search intent quickly and clearly
• Structuring headlines and sections so they’re easy to parse
• Breaking content into short, relevant questions that match how people search
• Adding original value, like data, examples and frameworks

How it scaled the work. Webflow’s blog has over a thousand articles, and its team could refresh about four a month by hand. It brought in a content automation tool to handle research and drafting, with human review checkpoints before anything was published. Webflow says it went from refreshing fewer than 50 articles a year to optimizing dozens a month, and saw a 40% lift in organic traffic to the updated content within days.

Worth taking from it. Webflow didn’t start by guessing at prompts. It started from the pages already working and asked why. Every SaaS company with analytics can do that this week.

Worth knowing. Webflow’s own conclusion is that optimizing for AI search is mostly SEO with a new lens, and still worth investing in.

The number. In June 2025, Patrick Stox, then on the technical SEO team at Ahrefs, published that over the previous 30 days AI search made up 0.5% of Ahrefs’ traffic but brought in 12.1% of its signups. He reported that AI search visitors converted at 23 times the rate of traditional organic search visitors, and that most of those signups came from ChatGPT.

How it was measured. Ahrefs’ own web analytics. Ahrefs counts AI search as search inside AI assistants like ChatGPT, Copilot and Gemini.

What Ahrefs says it did. On this one, straight answer: the post doesn’t explain why. Stox writes that he hadn’t yet looked into the cause, and wonders whether ChatGPT had simply fixed how it passes referral data. That’s a fair point for anyone reading their own numbers. Part of an increase can be measurement catching up.

What Ahrefs did share:

• Where AI visitors land. About 80% go to the homepage, the free tools and the product pages
• How they behave. On Ahrefs’ site they bounced less and viewed more pages than search visitors, but spent less time on site

What Ahrefs’ wider research says. Ahrefs has also studied this across other companies, and two findings matter here.

• Off-site mentions correlate most. In a study of 75,000 brands, Ahrefs found that the number of times a brand is mentioned across the web had the strongest link to showing up in Google’s AI Overviews. Mentions correlated far more strongly than backlinks did. Ahrefs is careful to say this is correlation, not proof of cause
• One company’s conversion rate isn’t a rule. Across about 82,000 websites, Ahrefs found that AI visitors were slightly more likely to bounce than search visitors. So the strong result on Ahrefs’ own site doesn’t automatically carry over to yours

Worth taking from it. Ahrefs’ number is the clearest example here of why AI search is worth measuring: a tiny share of traffic produced a large share of signups. And its honesty about not knowing why is the right posture for anyone reporting this internally.

The number. Tally’s co-founder, Marie Martens, publishes a milestone update every time Tally passes a revenue mark. In April 2026 she wrote that AI-powered search had overtaken Tally’s long-running word-of-mouth loop as its biggest acquisition channel. In her September 22, 2026 update, she wrote that 43% of new users now discover Tally through AI tools like ChatGPT, Claude, Gemini and Perplexity.

How it was measured. An onboarding survey that asks every new user where they found Tally. Martens says it also captures which prompts people used, the search terms that led to Tally, and which communities, creators and newsletters mention it.

What Tally says it did. Tally’s own account is the most specific of the four.

• The community comes first. Martens writes that word of mouth and AI search run on the same thing: happy users sharing Tally and recommending it on Reddit and elsewhere online
• It collects reviews steadily. Tally has been growing its presence on G2 and Capterra, asking for reviews in its newsletter and when closing support tickets
• It invests in AEO directly. Tally tracks where AI tools mention and cite it, looks for prompts where it’s missing, audits its help center to keep it accurate, and keeps its competitor comparison pages up to date
• It listens. Tally runs an NPS survey, an in-product feedback button and monthly open office hours, and shares every piece of feedback with the whole team

Worth taking from it. Tally didn’t buy this channel, and Martens notes it can’t be bought, at least not yet. What Tally did was build a product people recommend, then make sure the pages AI reads about it were accurate.

Worth knowing. Martens is clear that this is a channel Tally doesn’t fully control, and that a model update could reshuffle recommendations overnight. That’s why Tally keeps investing in the product, the free plan and the community.

Only the patterns at least three of the four stated themselves made this list. A ✓ means the company said it, in the source linked below.

1. They measured it first. Each had a way of counting: referrers, signup attribution, web analytics or an onboarding survey. You can’t manage what you can’t see, and none of them waited for a perfect method.

2. They care about what’s said about them elsewhere. Vercel seeds real mentions in developer communities. Webflow works on mentions in publications, Reddit and YouTube. Tally leans on its community and its reviews. Ahrefs’ research found that web mentions had the strongest link to AI visibility of anything it measured.

3. They make their own pages easy to understand. Vercel and Webflow both write for clarity and structure. Tally audits its help center and keeps its comparison pages accurate. If an AI tool reads your comparison page, it should come away with the right answer.

4. They keep it current. Vercel reviews on a 30, 90 and 180 day cycle. Webflow refreshes dozens of articles a month. Tally keeps its comparison pages and help center up to date.

5. None of them claims to have it figured out. Vercel says nobody has all the answers. Webflow says its work is just beginning. Ahrefs says it hasn’t looked into why, and that correlation isn’t causation. Tally says it doesn’t fully control the channel and a model update could reshuffle things overnight. That isn’t a weakness in their results. It’s how anyone credible talks about this channel.

Take these to whoever owns AI search. You’re not looking for perfect answers. You’re looking for which ones nobody can answer yet.

Measurement

• What share of our new signups came from AI last month, and how do we know?
• If we asked every new signup where they found us, what would the answer be?
• Are we counting AI referrals in our analytics at all, and which tools are we counting?

Being named

• Which buyer questions would we expect to be named on, and when did someone last check them in ChatGPT, Claude, Perplexity and Google AI Mode?
• When we’re not named, who is?
• When we are named, is what the AI says about us accurate?

What AI reads about us

• Where are we mentioned outside our own site, and is any of it out of date?
• Do our comparison and alternatives pages say accurate, current things about us and our competitors?
• Can an AI crawler read our key pages without running JavaScript?

Ownership

• Who owns this, by name?
• How often do we review the pages that matter most, and when did that last happen?

The two cheapest methods are the ones these companies used. Start with both.

1. Ask on signup. Add one question to your signup or onboarding flow: “Where did you first hear about us?” Include options for ChatGPT, Claude, Gemini, Perplexity and Google AI Mode, plus a free-text box. This is how Tally measures it, and it catches buyers your analytics never will.

2. Segment AI referrers in analytics. Create a segment for visits referred by AI assistants, for example chatgpt.com, perplexity.ai, claude.ai, gemini.google.com and copilot.microsoft.com. Then compare that segment’s share of traffic with its share of signups, the way Ahrefs did.

Time to set up: about 30 minutes for the signup question and about 30 minutes for the analytics segment. Then give it a full month before you read anything into it.

⚠️ Referrer domains change, and not every AI tool sends referral data. Check the list against your own referral report every quarter, and treat the analytics number as a floor.

It won’t tell you what will work for your company. These are four companies’ accounts of their own results. Your category, buyers and competitors are different.

It won’t prove why any of them grew. Each company describes what it did, and those descriptions are worth learning from. None of them claims to have proven which action caused which result, and neither does this.

It won’t stay current. These numbers are dated for a reason. Check each source before you quote it in a meeting.

It won’t put you in an AI answer. Nobody controls what a model says, including these four companies.

Every number and claim above, with its publication date. All checked on October 1, 2026.

• Guillermo Rauch (Vercel CEO), post on X, March 7, 2025. https://x.com/rauchg/status/1898122330653835656
• Guillermo Rauch (Vercel CEO), post on X, April 9, 2025. https://x.com/rauchg/status/1910093634445422639
• Kevin Corbett and Malte Ubl, Vercel, How we’re adapting SEO for LLMs and AI search, June 10, 2025. https://vercel.com/blog/how-were-adapting-seo-for-llms-and-ai-search
• Vivian Hoang, Webflow, How we’re optimizing for AI search at Webflow (and what we’ve learned so far), last updated February 24, 2026. https://webflow.com/blog/inside-aeo-strategy
• Patrick Stox, Ahrefs, Does AI Search Traffic Convert Better Than Traditional Search? For Ahrefs, Yes: 0.5% of Visitors Drove 12.1% of Signups, June 16, 2025. https://ahrefs.com/blog/ai-search-traffic-conversions-ahrefs/
• Louise Linehan, Ahrefs, An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied), May 26, 2025. https://ahrefs.com/blog/ai-overview-brand-correlation/
• Louise Linehan, Ahrefs, AI Visitors Visit Fewer Pages and Bounce More Often Than Traditional Search Visitors, June 24, 2025. https://ahrefs.com/blog/ai-traffic-quality-study/
• Marie Martens, Tally, The road from $4M to $5M ARR, April 15, 2026. https://blog.tally.so/the-road-from-4m-to-5m-arr/
• Marie Martens, Tally, 6 years in, $6 million far, September 22, 2026. https://blog.tally.so/6-years-in-6-million-far/

Everything above is free to use. The part most teams never get to is the measurement itself: the same buyer questions, checked the same way, month after month, in every engine your buyers use.

That’s what the AI Search Assessment is. $1,500, one time.

• Up to 20 buyer questions, built from what you know about your buyers and approved by you before anything runs
• Every one checked in ChatGPT, Claude, Perplexity and Google AI Mode: do you show up, who shows up instead, is what it says about you accurate, and which sources sit beside the answer
• A report card with the raw answer behind every row, in your hands a day before we sit down
• A 90-minute working session with me: 60 minutes on what to address first, 30 minutes training your team to run the check themselves
• A written plan within three business days, plus the recording and the transcript

It starts with a free 30-minute fit call. If it isn’t a fit, I say so and you owe nothing.

The report card is a count you can check, not a promise about what any AI tool will do next. I can’t promise an AI will recommend you. Nobody controls that.