How AlphaSense Wins 72.9% of AI Voice Share and the 4 Fixes You Can Copy This Week

Ask ChatGPT or Claude which tools lead market intelligence and AlphaSense is in most of the answers. A Catalyst scan in August put it at 72.9% of AI voice share in that category, against a visibility checklist score of 69 out of 100. The score is good and not remarkable. The share of voice is close to total.

An AI search teardown is a page-by-page read of a company that already wins AI answers, built for marketers who need to know which of those pages to copy and in what order. This one covers 6 AlphaSense pages, the homepage, the comparison hub, the ChatGPT comparison, the earnings pages, the blog and the pricing page.

The reason to study this one is that almost none of it depends on being AlphaSense.

Why does AlphaSense win AI answers when bigger budgets do not?

Because every page they win with answers exactly one question a buyer asks, and nothing else.

The clearest example is a page titled AlphaSense versus ChatGPT. They wrote a comparison against the language model itself. When a buyer asks an LLM why they should not just use ChatGPT for this, the answer it finds was written by AlphaSense. Nobody paid for that placement. They wrote the best page on that exact question before anyone else did.

"They're treating the AI tools as competitors, which they are." — Will Leatherman, Founder of Catalyst

That is the whole pattern repeated 6 times. A page per question, in plain words, where a machine can read it.

Is your category word somewhere a crawler reads first?

Check the title tag before you touch the headline, because that is where AlphaSense makes the match.

The AlphaSense homepage title tag reads "AlphaSense Market Intelligence and Search Platform". The category is named right there. The description underneath it mentions more than 500 million documents, so 2 short lines say what the company is and give a reason to believe it.

Then look at the headline, the biggest text on the page. "Accelerate your workflow with AI insights you can trust." It never says market intelligence anywhere.

"So AlphaSense gets the match from one small line most people never look at." — Will Leatherman, Founder of Catalyst

Most B2B homepages have the opposite problem. A clever headline that never says what the company is, and a title tag that repeats it. Headlines get written for people who already know you. An LLM deciding which companies belong in a category does not know you yet, so it needs the plain words.

Fetch the AlphaSense homepage without JavaScript and every headline still comes back, including "trusted by 7,000 of the world's largest enterprises". The structured data names both founders, the year they started and the address. That matters more than it sounds, because most AI crawlers other than Google's and Apple's do not run JavaScript. A site that builds its headings in JavaScript shows those tools a much emptier page than you see.

Your 4 minute version of this check is to open your homepage source and find whether your category word sits in the title tag, the headline, or neither. If the answer is neither, that is your first fix and it is a copy change. Then turn JavaScript off in the browser and reload. Whatever is still on screen is roughly what most AI tools get.

How many comparison pages should you have and which ones?

AlphaSense runs 41, and the list is every name a buyer in their market might put next to theirs.

The comparison hub carries pages against Bloomberg, PitchBook and Tegus, plus alternatives pages for Klue, Crayon, CB Insights, GLG and Third Bridge. Comparison questions are what buyers ask when they are close to buying, and a page titled with that exact matchup is the easiest thing in the world for an LLM to find and quote.

Two moves on that hub are worth stealing outright.

The first is the pages against the language models. AlphaSense has comparisons against ChatGPT, Perplexity and Claude, and since August they added a Copilot alternative page. Your buyers are asking whether they need you or whether a general model will do. Somebody is going to answer that question and it may as well be you.

The second is stranger. They publish pages comparing 2 of their rivals to each other, like CB Insights versus Crayon, with AlphaSense not in the title. Someone researching 2 other tools lands on alphasense.com while they are still deciding, which puts you in the conversation earlier than a versus page ever could.

The objection that comes up every time this gets raised with clients is that a comparison page promotes the competitor. If you know who your competitor is, your buyer almost certainly does too. Getting ahead of that question is the only version of it you control.

Does a comparison page need to admit where you lose?

Yes, and the ChatGPT page is the model for how to do it without giving anything away.

That page credits ChatGPT on several rows, including interface, cited summaries and finding key numbers. Some of those marks carry a condition next to them, like via plugins or enterprise only. It gives credit, adds the qualifier, and still steers. Further down there is a section on when ChatGPT is the right choice, which names general writing, coding help, and teams whose research mostly lives in their own internal documents.

"You'd think admitting weakness costs you. But a page that's honest about the other side gets trusted." — Will Leatherman, Founder of Catalyst

A page claiming to win every row is a page an LLM has no reason to believe. When a buyer asks for a balanced answer, the LLM wants a source that reads as balanced, and once it is quoting that page it is also quoting the rows where AlphaSense wins. One honest row buys the credibility of everything around it. The same mechanic sits underneath co-citation for AEO, where appearing next to credible sources does more for you than claiming credibility yourself.

The gap on that page is small and cheap. It carries a 6 question FAQ written as plain text, with no structured data labelling it as an FAQ. The questions are already written. The site just is not telling the LLMs that these are questions and answers, which is a few lines in a template using the FAQPage type, applied once, covering every FAQ on the site.

Ask one question of your own comparison pages. Is there a single row where the other side wins? If there is not, that is the row to add.

What is the strongest page you can publish that a competitor cannot?

One built on a number that only exists inside your company.

AlphaSense runs an earnings page for each of 171 public companies, with their own sitemap file pointing crawlers straight at the set. The AbbVie page summarises the last earnings calls in plain text and updates on the earnings calendar, so all 171 refresh on a schedule without anyone deciding to refresh them.

The part that cannot be copied is the sentiment score. It read +30 on the day of the teardown, down 11 from the previous reading, and the number comes out of an AlphaSense model. If an LLM wants to answer how the market felt about that quarter, there is exactly one place to get it.

"Opinions can be found everywhere on the internet, but your own data can only be found on your website." — Will Leatherman, Founder of Catalyst

This is the same move Ramp made on its vendor pages, and it has the same gap. The pages do not label the numbers with the Dataset type, so an LLM has to work out which numbers are on the page by itself. There are no dates in the structured data either, so the page refreshes every quarter and nothing in the code tells a machine when the numbers are from.

"The text says it and the labels don't." — Will Leatherman, Founder of Catalyst

Work out what your company counts that a competitor cannot. Usage data, transactions, benchmarks, support volume. It does not have to be 171 pages. One yearly benchmark from your own customers, dated, refreshed on a schedule, grows into the same asset.

What do dates and authors do for a blog post?

They tell an LLM who wrote this and how fresh it is, and AlphaSense wires both correctly.

Their resource posts carry a named author with a job title, a published date and an updated date, all in structured data, and the author name links to an author page for a real person with a real job rather than a company logo. Ramp's blog did not have this at all, and it was the first fix in that teardown.

The miss is in the copy. A post updated 2 days before the session still carried the previous year in its title, so the updated date in the structured data and the year on the page disagree. An LLM reading both has to pick one. When you refresh a post, move the year in the title and the headline at the same time.

The bigger problem on that side of the site is the resource hub itself. With JavaScript off it returns zero results, and there are more than 400 pages behind it. A crawler that cannot run a browser sees an empty list and never reaches any of them. That is one fix that unlocks the whole library.

Which page costs you the most by staying empty?

The pricing page, and AlphaSense leaves that one open.

The headline reads "Connect with our sales team for scalable pricing." The FAQ asks how much AlphaSense costs and answers please reach out. There is no number anywhere on the page.

When a buyer asks an LLM what AlphaSense costs, it cannot get an answer from AlphaSense, so it takes one from a review site, a forum or a competitor's comparison page. Those sources may be stale, wrong, or written by somebody with a reason to inflate the figure, and the buyer walks away believing they have the answer.

"So the pricing page tells the AI everything about AlphaSense except the one thing that buyers actually need to make a decision." — Will Leatherman, Founder of Catalyst

The irony is that the same page is the most detailed one on the site about who they are. Founded in 2011. More than 7,000 customers. Most of the S&P 100. Over 500 million documents. Everything except the price.

Enterprise deals genuinely do vary, so there is a middle ground that most legal teams will sign off. A starting price, a typical range for a team of a given size, or an honest explanation of what moves the number. Any of those gives an LLM something to quote. Ask ChatGPT what your product costs and look at where the answer comes from. If it is not from you, a range on your own page fixes it.

Which 4 fixes should you run first and who owns each one?

Cheapest first, hardest to approve last.

Every one of those is something AlphaSense already does well somewhere else on their own site. None of them is a budget problem.

Does a high visibility score explain why a company wins?

No, and this is the honest limit on everything above.

"We've seen what the leader ships. We haven't proven that any one of these things is why they win." — Will Leatherman, Founder of Catalyst

A Catalyst scan of Ramp in June scored it far below AlphaSense on the same checklist, and Ramp still led its category on AI voice share. A checklist score is not the whole story. What both companies have is pages that answer the question their buyer is actually asking. Test the fixes on your own site rather than taking them on faith, which is what the 20 minute AI search audit is for.

One more thing worth knowing, since it comes up constantly. AlphaSense has no llms.txt file at all. Requesting it returns nothing, and they are winning their category anyway. Do not lose sleep over that one.

The takeaway

AlphaSense holds 72.9% of AI voice share in market intelligence with a 69 out of 100 visibility score, 41 comparison pages, 171 data pages that refresh themselves, and a pricing page that answers nothing. The pattern behind the wins is one question per page, answered in plain words a machine can read.

Do 3 things this week. Open your homepage source and check whether your category word is in the title tag, the headline or neither. Write down the 3 versus questions your sales team hears most, then add a fourth, you against ChatGPT. Next week, write the first one, with one row where the other side genuinely wins.

If you want the prompts and page templates behind this, the AEO skills pack has them, and a free AI visibility audit will tell you where AI answers currently put your company and what to fix in what order.