gshc2020.com

But here’s the silver lining: You can be invisible in blue-link results and still win AI visibility if you know how to optimize for it.

I’ve spent a lot of time inside the data, working with content marketers who are trying to figure out exactly this. What follows is everything that actually works, backed by research — no fluff, no guessing.

Table of Contents

• Why Ranking in AI Search Results Matters
• How to Optimize Content for AI Search Prompts
• How to Track AI Search Ranking Performance
• How to Get Started with AI Search Rank Optimization: A 3-Month Plan
• How to Rank in AI Search Results with HubSpot Solutions
• Frequently Asked Questions About AI Search Ranking

Why Ranking in AI Search Results Matters

AI-referred visitors aren’t just traffic — they’re pre-qualified buyers. Ahrefs analyzed their own traffic data and found that AI search visitors accounted for just 0.5% of total visitors, but drove 12.1% of all signups. That’s 23x the conversion rate of visitors from traditional organic search.

Semrush confirmed the pattern, finding that AI search visitors on average convert at 4.4x the rate of standard organic visitors.

These visitors are inherently more qualified than visitors forced to click onto your website. In most instances, they’ve already received the answer they needed from AI and actually chosen to click through for more. That self-selection shows high intent and interest.

The volume of AI search traffic is still small compared to Google, but it’s growing fast. The teams investing in AEO (or GEO) now are building citation authority while competition is still low. If SEO trends continue, that tide won’t stay low for long.

How to Optimize Content for AI Search Prompts

1. Make sure your website is accessible to AI crawlers.

Cloudflare reported that AI crawlers now account for 4.2% of all HTML requests across their network. OpenAI’s GPTBot alone grew 305% from May 2024 to May 2025. However, if your robots.txt or server configuration blocks AI crawlers, even the greatest content will go unnoticed by AI knowledge bases.

Seasoned digital marketers know that before search engines can even rank you, they need to be able to crawl your pages, and it’s the same for AI search results. Every major AI platform has its own crawler.

Here are the most important ones to know about:

Platform	Crawler / User-Agent	Purpose
ChatGPT Search	OAI-SearchBot	Real-time retrieval (not training)
OpenAI	GPTBot	Model training
Perplexity	PerplexityBot	Real-time retrieval
Anthropic / Claude	ClaudeBot	Training and retrieval
Google AI Overviews	GoogleBot	Indexing and retrieval

Now, if you have some sort of intellectual property (IP), private, or proprietary content on your website, you don’t want AI using without compensation, having AI crawlers blocked isn’t a bad thing — but make sure you’re not blocking crawlers you want to let in.

Check your robots.txt file.

OAI-SearchBot and PerplexityBot, for example, are retrieval crawlers. They don’t use content for training, but they power real-time AI answers. Block them, and you disappear from ChatGPT and Perplexity search results.

Pro Tip: Even if you’re protective about your content, don’t block all AI crawlers. Research from Rutgers Business School and Wharton found that publishers blocking AI crawlers via robots.txt lost roughly 7% of weekly traffic within six weeks.

AI crawlers aside, there are several other things you can do to make your website technically accessible to AI.

Consider adding an llms.txt file.

An llms.txt file is a document added to your website that serves as a map and resource guide for AI models, search agents, and autonomous web bots. This newer standard, officially supported by Anthropic, helps AI systems understand which content is safe to summarize and cite.

What we like: The llms.txt file is a quick win. It takes under an hour to create and tells AI crawlers clearly which parts of your site you want them to use. Think of it as a welcome mat for AI systems.

Improve your page speed.

AI bots prioritize fast servers (and let’s face it, users want fast sites too). Aim for sub-200ms TTFB (Time to First Byte) to ensure your content is crawled frequently and refreshed quickly. Use HubSpot to check your site speed, then try these ways to improve your page loading speed.

Fix crawl errors.

Broken pages (404s), redirect chains, and invalid sitemaps can also reduce crawl budget. So, keep your robots.txt error-free. Google Search Console is great for catching technical errors that prevent both Google and AI crawlers from reading your content.

Make sure Bing indexes you.

Google is still the top search engine, but ChatGPT Search is built on Bing. If you’re not indexed on Bing, you may not appear in ChatGPT search results. Set up Bing Webmaster Tools and submit your sitemap.

2. Lead with an answer-ready (or answer-first) structure.

AI systems don’t read content the way humans do. They scan for easy-to-extract answers to users’ queries and intents. If your page doesn’t make these answers obvious, the AI skips to a page that does.

Answer-ready content starts with a direct answer. And this is not a suggestion — it’s the single most reliable structural tactic for AI citation. Let’s get more granular.

• Use question-based headers (H2s and H3s). Structure sections around the natural-language questions your audience types into AI search. Think “How do I...” and “What is...”
• Lead with the answer. Don’t bury the answers to queries three paragraphs down. State it clearly in 1-3 sentences using plain language. Aim for 75–150 words.
• Use short sentences. AI systems favor content written for human comprehension. Long, complex sentences get paraphrased badly or skipped entirely.
• Support it with facts. Follow with data, examples, or context that proves the point.
• Write for the query, not the topic. AI search is conversational. Your content needs to match the way real people phrase questions, not how marketers write about topics.

TL;DR: Whatever your heading promises, deliver it immediately. Don’t make the reader wade through context before getting the answer.

Example Header: “How Does Content Marketing Drive Revenue?”

STRUCTURE Type	EXAMPLE
Before answer-led	"In today's competitive digital landscape, brands are increasingly looking for ways to connect with their audiences in more meaningful ways. Content marketing has emerged as one of the most discussed approaches ..."
After answer-led	"Content marketing drives revenue by attracting high-intent visitors through search and converting them with useful content before they ever talk to sales. Companies that blog consistently generate 67% more leads per month than those that don't."

Pro tip: Use HubSpot Content Hub’s AI writing tools to restructure existing blog posts into an answer-first format. Paste your section into the AI editor with the prompt: “Rewrite this to lead with a direct 2-sentence answer to [question].” It takes minutes per section.

Depending on the nature of your page or section, you may also want to use schema markup in your page structure or, more specifically, FAQ schema. More on that in our next section.

3. Use structured data on your pages.

Structured data or schema markup is one of the most effective ways to translate and communicate your content value for AI. Schema outlines the meaning of your content explicitly. Without it, AI systems have to guess based solely on what’s visibly on the page, and let’s face it, clarity isn’t every brand’s strong suit.

But that doesn’t make schema a magic wand for AI search either.

Three important things to remember:

• Schema is only confirmed to increase your potential for Google AI Overviews. Its impact on ChatGPT, Perplexity, and Gemini is still unclear.
• Schema markup doesn’t directly affect rankings. Google’s own Search Central documentation states explicitly that structured data changes “won’t affect how pages are ranked.” But it significantly improves how AI systems understand and extract your content.
• For ChatGPT and Perplexity, visible on-page Q&A formatting matters most. These LLMs read your JSON-LD as raw text. Your schema signals intent to Google’s Knowledge Graph, while the visible content structure is what ChatGPT and Perplexity extract directly.

There are several types of schema markup, but here are the ones most likely to improve AI search performance and are applicable to most businesses.

Schema Type	What It Does	Best For
FAQPage	Signals Q&A content structure to Google AI	Blog posts, help articles
Article	Identifies author, date, and topic for content clarity	All editorial content
Organization	Confirms brand identity and contact details	Homepage and about pages
HowTo	Structures step-by-step instructions	Tutorial and guide content
Product	Defines product details, pricing, and reviews	Product pages

When implementing these, always use JSON-LD format. It’s cleanly separated from your HTML, makes it easier for AI crawlers to parse, and is explicitly recommended by Google. Then, validate your schema with Google’s Rich Results Test and fix any errors before publishing.

Data and comparison tables also help. Like schema, tables organize your data in an easy-to-understand way. If your content compares options, shows data, or lists features, put it in a table.

4. Organize your content into pillar pages and clusters.

Topic clusters build topical authority. AI systems use topical authority signals to decide which sources to trust on a given subject, and you want your business or brand to be one of them.

A site that covers a topic deeply (say with a pillar page and supporting cluster content) signals expertise that a single blog post can’t.

Fan-out is why this matters even more now. Fan-out is the process by which an AI system takes a user query and breaks it into multiple related sub-queries before generating an answer.

For example, if someone asks ChatGPT: “What’s the best CRM for a small sales team?”

The AI doesn’t just search for that exact phrase, but expands into:

• “CRM software comparison for small businesses”
• “best CRM for sales teams under 20 people”
• “HubSpot vs Salesforce for small teams”
• “affordable CRM with pipeline tracking”

This is important because it means your content can still be cited even if you don’t rank for the initial query, as long as you cover the subtopics on your cluster pages well. The subtopics show you go deep on a topic; you don’t just scratch the surface.

The internal links created by clusters are also essential. When you link from a cluster page to your pillar and vice versa, you’re creating a semantic web that AI systems can follow. It shows them which content on your site is most authoritative on a given topic.

Learn how to get started with pillar pages.

Pro tip: Use HubSpot’s SEO tool to identify content gaps in your cluster. It maps your existing content to topic themes and shows you which subtopics are missing — so you know exactly what to create next instead of guessing.

5. Follow Google’s E-E-A-T framework.

E-E-A-T (Experience, Expertise, Authoritativeness, and Trustworthiness) isn’t just a Google quality signal anymore. It’s an AI citation filter. But E-E-A-T isn’t about writing style, it’s about proof. AI systems look for signals that the content comes from a real, credible person with real experience.

Add the following to your website:

• Author bios with credentials. Make it easy for AI to verify who wrote the content and that they actually know what they’re talking about. Include their name, job title, years of experience, photo, and links to published work.
• First-person experience. Anecdotes from real practitioners outperform generic advice. That said, use phrases like “I tested this” or “In my experience...” when you have direct experience.
• Data citations. Link to reliable, primary sources. This is a big quality signal.
• Original research. Proprietary data and survey results have no substitutes. AI systems must cite the original source. Original research is the #1 type of content for earning AI citations.

6. Optimize for authority off-site.

On-site content only gets you so far. Think of it like someone telling you they’re the best chef in the country. You certainly wouldn’t just take their word for it; you’d want to confirm with third parties. Authority works the same way with AI systems.

Brands are 6.5x more likely to be cited by AI via third-party sources than via their own domains, according to Airops 2025.

To build off-site authority:

• Earn mentions in reputable publications. Guest posts, expert quotes, positive mentions, and PR placements in well-known industry publications all vouch for your authority in your area of expertise.
• Maintain consistent brand data. Inconsistent contact and brand info confuses AI and buyers alike. Make sure your name, address, phone number, product details, and descriptions match across all your internet presences (LinkedIn, G2, Crunchbase, Wikipedia, and industry directories).
• Monitor AI misrepresentation. Run brand queries in ChatGPT, Perplexity, and Google AI regularly. If AI is misrepresenting your brand, prioritize publishing content that crowds out the misinformation.
• Get active on Reddit and in communities. Brand mentions in discussions weigh just as mentions on social media do.

Pro tip: Create a Wikidata entity for your brand if you don’t have one. A clean Wikidata entry is one of the fastest E-E-A-T wins for ChatGPT visibility. It gives AI systems a machine-readable source of verified brand facts.

7. Refresh your content regularly.

Content refresh timing depends on the topic and how fast the space is moving. Here’s a simple framework:

Content Type	Recommended Refresh Cadence
Pillar pages/cornerstone content	Every quarter
Blog posts with statistics	Every 6 months, or when key stats are outdated
Product/feature pages	Within 30 days of any product change
FAQ sections	Every 3 months, based on new customer questions

Pro tip: When you refresh your content, update the publish date.

AI Overviews and RankBrain favor recently updated content. So, a page refreshed in March 2025 will outperform an identical page last updated in 2022, even if the actual content is similar.

How to Track AI Search Ranking Performance

You can’t improve what you don’t measure, but AI citation tracking is truly different from traditional rank tracking. Here’s the framework I recommend for building a real AI visibility measurement practice.

There are three metrics that matter most for AI visibility:

Metric	What It Measures	How to Track
Citation presence or Visibility	Does AI mention your brand/content in answers?	HubSpot AEO, Otterly.AI, Semrush AI Toolkit
Share of voice	How often do you appear vs. competitors in AI answers?	HubSpot AI Search Sensor, manual brand queries
AI-referred traffic quality	Are AI-sourced visitors converting?	GA4 session source, CRM attribution

These are new metrics for most, of course, so you likely don’t have baseline data to set goals or even evaluate performance.

To set your baseline:

• Run 20-30 target queries in ChatGPT, Perplexity, Google AI, and Claude. Record whether your brand is cited. You can also get a snapshot of your performance with HubSpot’s free AEO grader.
• Track your AI-referred sessions in Google Analytics 4 under source/medium. Look for referrals from ChatGPT, Perplexity, and other AI platforms.
• Benchmark your citation rate against your top three competitors for the same queries.

From there, set a 90-day target. AI search optimization typically yields initial results within 2-3 months of implementation.

What we like: HubSpot AEO tracks your AI citation presence and brand mentions across all major AI platforms, gives you a readiness snapshot with the AEO Grader, and benchmarks your performance against industry trends via AI Search Sensor. It connects AI visibility directly to your CRM data so you can see which AI-referred visitors actually convert. We’ll get deeper into that in a few sections.

How to Get Started with AI Search Rank Optimization: A 3-Month Plan

Start with what moves the needle fastest and builds a foundation for everything else. Here are the prioritized steps I recommend for this quarter:

1. Technical access (Week 1). Audit robots.txt, fix blocked AI crawlers, and confirm Bing indexing. This is a must for everything else.
2. AEO baseline (Week 2). Run HubSpot’s free AEO grader and target queries across ChatGPT, Perplexity, and Google AI. Document your current citation rate.
3. Schema implementation (Week 2-3). Add JSON-LD Article and FAQPage schema to your top 10 most-visited pages.
4. Answer-first refresh (Week 3-4). Rewrite the intros of your top 20 pillar pages to lead with direct 1-3 sentence answers.
5. Content cluster audit (Month 2). Map your existing content to topic clusters. Identify the biggest gaps and fill them.
6. E-E-A-T and off-site authority (Ongoing). Build a cadence of original research, expert guest posts, content updates, and brand mention monitoring.

→ See how HubSpot connects AI citation tracking to CRM pipeline — Get a demo

How to Rank in AI Search Results with HubSpot Solutions

HubSpot has built AI visibility and AEO tools directly into our platform, so you can operationalize most of this program without adding new software.

Here’s how HubSpot’s tools map to each part of the AI search ranking program:

HubSpot Tool	What It Does for AI Search
HubSpot AEO	Tracks AI citation presence, brand mentions, and visibility across platforms. Connects AI referrals to CRM pipeline.
AEO Grader	Gives you an AI readiness score for any page or domain. Flags structural, schema, and content issues with recommendations.
AI Search Sensor	Tracks industry benchmarks and AI citation volatility. Tells you how your share of AI voice compares to competitors.
Content Hub	Manages topic clusters, pillar pages, and internal linking at scale. AI writing tools help restructure content for an answer-first format.
Breeze AI	Automates content refresh suggestions, identifies outdated stats, and recommends AEO improvements across your content library.
Smart CRM	Attributes AI-referred sessions to contacts and deals, so you can see which AI channels are actually driving revenue.

Now, the real advantage of managing AI search visibility inside HubSpot is attribution. Most teams track AI citations as a vanity metric, like “we appeared in 40 AI answers this month.”

HubSpot’s Smart CRM connects those citations to real results, such as sessions, contacts, deals, and revenue. That’s the difference between reporting on visibility and proving business impact.

Start with the AEO Grader. Run your top 5 pillar pages through it before doing anything else. It’ll show you exactly where the biggest gaps are, so you can prioritize your next steps.

Frequently Asked Questions About AI Search Ranking

How long until AI citations improve?

Brands with established topical authority and active content distribution can see citations improve within weeks of a major content refresh. Patience plus consistency is the formula.

Do I need separate strategies for Google AI Overviews and ChatGPT?

Yes, but there’s a lot of overlap between the core principles. A universal foundation of technical access, answer-first structure, schema, and E-E-A-T supports all platforms simultaneously.

The biggest differences:

• Google AI Overviews favor E-E-A-T signals, mobile optimization, and freshness. Organic ranking still helps, but it’s not required.
• ChatGPT incorporates Bing indexing and favors Reddit and Wikipedia in its training data. Active community presence matters more here.
• Perplexity favors conversational, experience-based content with practical examples and cited sources.

What if AI misrepresents my brand?

Unfortunately, this happens more than most brands realize. The fix is to be proactive, not reactive. Publish accurate, authoritative content that crowds out misinformation.

Make sure your brand facts are consistent across all platforms (e.g., website, LinkedIn, G2, Wikipedia, Crunchbase). AI systems learn from the most authoritative available sources. So,if you own those sources, you own the narrative. In critical situations, you can also consider blocking AI crawlers.

Should I block AI crawlers?

In most cases, no. If AI can’t crawl your site, it can’t cite it. The exception is if you have proprietary content or significant server cost concerns, you can selectively block training crawlers (like GPTBot) while allowing retrieval crawlers (like OAI-SearchBot and PerplexityBot) that power real-time AI search answers.

What’s the best way to track AI citations over time?

Build a simple tracking cadence with tools like HubSpot AEO. That may look like:

• Weekly manual checks: Run your 10 most important queries across major AI platforms.
• Monthly tool review: Use HubSpot AEO or Semrush AI Toolkit to track citation trends at scale.
• Quarterly attribution analysis: In GA4, review AI-sourced sessions and tie them to conversion events in your CRM.

AI citation rates fluctuate. Don’t panic about week-over-week swings. Look for the 90-day trend.

AI search is not the future; it’s our current reality. AI overviews, ChatGPT, Perplexity, Gemini, these platforms are where your audience is finding answers, and they’re usually finding them without clicking to your website.

The marketers who figure this out first will build durable AI visibility advantages that compound over time, while others spend the next two years trying to catch up.

You have the playbook. Now, run with it.

%!s()

Those acute hazards, which hang over 79% of capacity, include severe climate-induced events such as flooding, extreme winds and wildfires that can disrupt operations, increase downtime and drive insurance and repair costs, according to the report by First Street, a climate risk analytics firm.

First Street looked at 97 global data center markets in the report.

"Most underwriting for real assets still uses historical data, but the climate is no longer behaving the way the historical record would predict," First Street CEO Matthew Eby said in a release. "As heat, drought, and water stress increase, outdated models simply don't offer a complete view of risk anymore."

The study also found that just over half of all data centers globally are in markets exposed to chronic climate stress, such as extreme heat and drought, that hit energy efficiency and increase costs.

While risk can be heightened by exposure to singular climate-driven events, such as strong hurricanes, it is the chronic effects of climate change that can cause the most damage, both physically and financially.

Jeremy Porter, chief economist at First Street, said the backward-looking models are not correcting for climate and for all the sources of risk.

"Ultimately, it's just something that we're underestimating," he said.

Porter pointed to government models that he says are outdated because they look at past levels of precipitation and do not factor in the effects of climate change. As the Earth warms, clouds hold more moisture and rainfall becomes heavier.

The danger, according to the study's findings, is that investors may be looking at traditional metrics when developing data centers and not taking into account that climate could impact long-term operating conditions. Data centers are typically expected to be in operation for 20 to 30 years.

"Investors who incorporate these factors into underwriting and capital allocation decisions will be better positioned to identify resilient markets and avoid mispriced risk," Eby said in a release.

Get Property Play directly to your inbox

CNBC's Property Play with Diana Olick covers new and evolving opportunities for the real estate investor, delivered weekly to your inbox.

Subscribe here to get access today.

Some developers and operators are, however, taking climate into account as they design data centers.

Digital Realty, for instance, is working to make sure it has enough water to cool its centers.

"Almost all of our data centers today, 300 around the world, the global portfolio, are either a waterless system or closed loop water," said Andrew Power, CEO of Digital Realty on the Property Play podcast earlier this year. "So think of it as there's no evaporation. We make the investments and elect to do that."

While developers can implement resilient measures for building envelopes — the physical barrier between a building's interior and the outside — to protect structures from severe weather events, it's the actual systems that connect to the data centers that are most important in terms of understanding both internal and external vulnerabilities to climate risk, according to Porter.

He said developers need to look beyond building and parcel-specific risk to a systems-level thinking. That means understanding infrastructure and community demographics.

"Acute climate risk can be mitigated with building adaptation, so that, that's not a huge issue. Then you start to think about the system, though, you start to think about the infrastructure, the egress, the access to the site, power access. There is a building mitigation process, but then there's a community mitigation process," he said.

Some areas are more prone to climate risk than others, the study found.

The Asia-Pacific region has 89% of its data center capacity at risk, the highest percentage of any geographic area, according to First Street. That compares with 50% exposure in the Americas and 46% in Europe, the Middle East and Africa.

Some of the industry's fastest-growing markets are also the most exposed, the study found. That includes Northern Virginia in the U.S., as well as Johor in Malaysia and Marseille, France. Nordic markets had the lowest climate risk.

"The top 10 markets with acute climate risk across the globe, most of them are in the U.S. Most of them have wind and flood risk," said Porter, adding that the U.S. does not see as much chronic climate risk as the rest of the world.

watch now

%!s()

Stonehenge is also aligned to the summer solstice, and the wooden monument may have been an early prototype of this. It is one of the earliest examples of a monument aligned to an astronomical phenomenon in the British Isles.

“What we have now, for the first time, is actual proof that these people were capable of capturing the movement of the sun,” said Phil Harding at Wessex Archaeology, who led the excavations, at a press conference announcing the discovery.

Stonehenge is a monument built during the Neolithic, the very end of the Stone Age. Situated on Salisbury Plain in the UK, it consists of an outer ring of vertical sarsen stones topped by horizontal lintels, a smaller inner ring of vertical bluestones and several other stones. These are surrounded by an earth bank and a ditch. This is the oldest part, built around 3100 BC, with the stones being placed over the centuries up to 1600 BC.

Some of the standing stones seem to have been carefully aligned to point to the spot on the horizon where the sun rises on the summer solstice, and to the opposing spot where the sun sets on the winter solstice. The relevant stones were erected about 2500 BC.

Harding and his colleagues have found the remains of a nearby monument that is 500 years older. About 5 kilometres north-east of Stonehenge, there is a village called Bulford, where the UK’s Ministry of Defence wanted to house about 5000 army personnel. Before construction began, Wessex Archaeology excavated Bulford from 2015 to 2017.

The team found a cluster of pits containing grooved ware pottery, which was made by late Neolithic peoples. Radiocarbon dating suggested the pottery was from about 2950 BC. The researchers obtained 40 dates, all tightly clustered in time. “This site was being occupied for a relatively short period of time,” said Harding. “It could be something like a decade.”

“It’s a really important Middle Neolithic settlement,” says Susan Greaney at the University of Exeter, UK, who wasn’t involved in the study.

Two of the Bulford pits were a different shape to the others. Instead of having vertical sides, they tapered towards the bottom, going from about 1.2 metres across to just 0.5 metres. They also didn’t contain pottery, but were filled with chalk rubble. Harding and his colleagues concluded that they were postholes: they had once housed timbers a few metres tall, which were held upright by the rubble. In line with this, one of the postholes contained charcoal from an ash tree.

The two postholes are about 120 metres apart. Harding realised that a line drawn through them would point roughly north-east, at 48.1 degrees: about the direction of the midsummer sunrise. “I got really, really excited about that,” said Harding.

Wessex Archaeology recruited Fabio Silva, a skyscape archaeologist at consultancy Stone x Sky, to study the alignment of the postholes more carefully. Using a 3D reconstruction of the landscape with modern buildings removed, plus data on the sun’s shifting path across the sky, Silva determined that the postholes were neatly aligned with the past summer solstice sunrise.

Strictly, the alignment was about 1 degree out, but Silva said this makes sense once you realise that the wooden posts could have been up to 50 centimetres across. “You have to take that width into account,” he said at the press conference, in which case the alignment is “bang on”. “The odds of this being by chance are less than 0.5 per cent,” he said.

“Probably a rough orientation is good enough for the ritual that you are supposed to carry out in these sites,” says A. César González-García at the Spanish National Research Council in Santiago de Compostela, who wasn’t involved in the study. “It looks like there is a broad understanding and interest in the sky.”

Older sites in the area also show evidence of people tracking the sun, albeit with less precision. “From the earliest times that we have Neolithic people present in that landscape, they are incorporating the sun into their ceremonial architecture,” said Matt Leivers, also at Wessex Archaeology.

“We’ve got loads of timber monuments that have those kind of alignments,” says Greaney. The Bulford monument “is adding another one to that, potentially, but much earlier”.

For instance, at nearby Larkhill, there is a Neolithic enclosure from around 3700 BC, well before Bulford and Stonehenge. Its entrance faces roughly north-east. If you stand in that entrance and look at Sidbury Hill (the highest point on the horizon) on midsummer morning, “you see the sun rise dead ahead of you over Sidbury Hill”, said Leivers. %!s()

“The results are stunning,” says Peter Sasieni at Queen Mary University of London. “It’s an awful thing when somebody dies very young from cervical cancer. This is a real triumph for vaccination, a real triumph for science and a real triumph for public health to get that vaccine out there with very high uptake very rapidly.”

HPV is spread by several kinds of sexual activity, and many strains genetically modify cells in a way that is extremely likely to cause cancer. Women can get cervical cancer in their twenties because of the virus, and around the world many are still dying because of it.

The first HPV vaccine became available in 2006. In the UK, it has been offered to girls aged 12 or 13 since 2008. It’s been offered to boys since 2019, both to protect them from other HPV-triggered cancers – like those affecting the mouth, anus, throat and penis – and to prevent them from infecting others.

The study is the first evidence that the HPV vaccine prevents cervical-cancer-related deaths, in addition to dramatically reducing HPV infections and the incidence of cervical cancers. It may seem obvious that it prevents deaths, but we know that the women who are least likely to get vaccinated are also the least likely to go for screening, says Sasieni. So, there was a worry that the vaccine has mainly been preventing cancers that would have been detected early by screening and successfully treated, but not preventing those that would get missed by screening and therefore be more likely to kill women.

Fortunately, this doesn’t appear to be the case. Sasieni and his colleague Milena Falcaro, also at Queen Mary, have been monitoring cervical-cancer rates and deaths in England, and noticed that between 2020 and 2024 – the most recent data available – there were no deaths among women aged 20 to 24. Based on historical rates, around 23 deaths would be expected. “As far back as I’ve seen data, there’s never been a year with none, and so five years in a row with no deaths is really quite something,” says Sasieni.

The dramatic drop is almost certainly due to the HPV vaccine – around 90 per cent of women in England aged 20 to 24 were vaccinated when they were 12 or 13. “This hugely encouraging news shows the life-saving impact of the HPV vaccine, and it’s incredibly exciting to be able to say to this whole generation: cervical cancer and some other cancers shouldn’t be a risk for you,” says Caroline Temmink, director of vaccination for the National Health Service (NHS) in England.

While this study looked only at cervical cancer, the benefits of this vaccine also apply to other HPV-related cancers. It should also protect against warts on the skin and genital and anal areas in both sexes.

Sasieni and Falcaro note that there were some deaths among women aged 25 to 29, but far fewer than would be expected. Altogether, they estimate around 200 lives have already been saved, and this is just the start. “The 200 which we’ve estimated in the paper is really just the tip of the iceberg, because it looks like there’s long-term protection against infection,” says Sasieni. “In the future, 18,000 deaths might be a rough estimate of what we’re preventing so far.”

Worldwide, however, uptake of the HPV vaccine is low and the rate of cervical cancer is still rising. Death rates may also rise in the UK because fewer teenagers are getting vaccinated. “The bad news is that the vaccine uptake has fallen quite dramatically since covid,” says Sasieni.

“Alongside cervical screening, HPV vaccination is central to the NHS ambition to eliminate cervical cancer by 2040,” says Temmink. “It’s a safe and effective vaccine and we urge everyone eligible to take up the offer when invited.” %!s()

Researchers have spent decades investigating ways to chill people’s brains after they have a stroke to try to limit the damage. The idea is to freeze brain cells in a hibernation-like state so they don’t need as much oxygen and glucose when a stroke cuts off their blood supply. If brain cells can be kept alive until blood flow is restored, for example, by removing a clot, a patient may be spared from extensive brain damage and related speech and movement problems.

Unfortunately, the physical cooling strategies that have been looked at so far – including cooling blankets, ice packs and helmets – haven’t worked well. This is because they cause intense discomfort and uncontrollable shivering, says Kirsten Coupland at the University of Newcastle in Australia, who wasn’t involved in the study.

Shivering is one method the body uses to “fight the induction of hypothermia”, making it hard to get body temperature down low enough, she says. “It’s great to see different cooling therapies being tested out for stroke because we know that physical cooling is just not feasible,” says Coupland.

Shuaili Xu at Capital Medical University in Beijing, China, and his colleagues administered the two drugs, promethazine and chlorpromazine, which have been known to reduce body temperature since the 1950s, to mice and rhesus monkeys following induced strokes.

In both animals, the drug combination led to drops in core body temperature, suppressed glucose metabolism in cells and lessened the amount of brain damage caused by the stroke. The reduced brain damage also meant the treated monkeys displayed better use of their limbs.

Next, the team conducted a clinical trial involving 32 people who had just had a stroke. Upon hospital admission, the subjects were given the promethazine and chlorpromazine combination or a placebo, in addition to standard clot-removal therapy.

The promethazine and chlorpromazine treatment only reduced the patients’ body temperature by 0.3°C (about 0.5°F) and did not reduce stroke damage. However, Xu thinks this is because the infusions were done over 12 hours, which was too slow to bring down core body temperature by a meaningful amount. “It might have led to a low blood drug concentration per unit of time,” he says.

His team is now launching another trial to see whether faster infusions over an hour produce stronger cooling effects and therapeutic benefits. “The fact that they’ve proven that it’s safe and these drugs are already used in humans for other indications means I think that it’s reasonable to proceed with further clinical trials,” says Coupland.

Promethazine and chlorpromazine are known to be relatively safe because they have been used for decades. Promethazine is a sedating antihistamine that can ease hay fever and assist sleep, while chlorpromazine is an antipsychotic drug used in the management of schizophrenia and bipolar disorder. They both act on the central nervous system to reduce core body temperature, without translating into shivering or subjective feelings of coldness. %!s()