RELX, LSEG and Experian: Is the AI Sell-Off in UK Data Stocks Overdone?
· Members
Early this year, Anthropic posted a set of 11 plugins for its Claude AI on GitHub, one of them aimed at corporate legal teams. When Bloomberg reported on them, London's information giants fell off a cliff. According to Proactive Investors' live coverage, RELX, owner of the LexisNexis legal database, was down almost 17% late that afternoon. LSEG, the London Stock Exchange Group, was down 12.7% and credit bureau Experian 10%. Thomson Reuters fell 10.8% in New York and Dutch peer Wolters Kluwer nearly 10% in Amsterdam.
Watch it play outOne afternoon took a sixth off RELX
Picture RELX, owner of the LexisNexis legal database, as 100 dots of stock market value.
Early this year, Anthropic posted a set of 11 plugins for its Claude AI on GitHub.
One was aimed at corporate legal teams, the customers LexisNexis serves. Bloomberg reported on them.
By late that afternoon RELX was down almost 17%: about a sixth of the company, gone.
The rest of the sector fell with it: LSEG down 12.7%, credit bureau Experian down 10%.
Source: Proactive Investors live coverage
That afternoon was one step in a long slide. Measured from their highest weekly close, RELX is down 37%, LSEG 31%, Experian 38%, Rightmove 43%, Auto Trader 48% and Wolters Kluwer 60%. The FTSE 100, London's index of its 100 largest listed companies, sits just 4% below its own record.
Watch it play outThe data stocks lost a third or more while the FTSE 100 held near its record
Line up the FTSE 100 and three data stocks, each measured from its highest weekly close.
The FTSE 100, London's index of its 100 largest companies, is just 4% below its record.
RELX is down 37% and Experian 38%. LSEG has lost 31%.
Auto Trader is down 48% and Rightmove 43%. Wolters Kluwer has lost 60%.
Source: Yahoo Finance weekly closes; Yield Theory calculations
The odd thing is what the businesses did meanwhile. RELX's half-year results showed 7% underlying revenue growth and adjusted earnings per share up 11% at constant currency. Its share price fell anyway, so the price-to-earnings ratio, or P/E, the price divided by a year's earnings per share, has collapsed. At the peak, investors paid about 34 times earnings. Now they pay about 19.5 times, 43% less per pound of profit.
Watch it play outRELX's earnings grew, yet the multiple fell 43%
A share price is earnings per share (across) times the P/E (up). At the peak, RELX's P/E was about 34.
Since then, adjusted earnings per share rose 11% on 7% revenue growth. The base widens.
But the share price fell anyway, so the P/E collapsed to about 19.5 times.
More profit, cheaper shares: investors now pay 43% less for each pound RELX earns.
Source: RELX results; Yahoo Finance; Yield Theory calculations
This piece tests the AI fear against what eight companies actually reported: RELX, LSEG, Experian, Wolters Kluwer, Sage, Pearson, Rightmove and Auto Trader. It covers their growth, margins and buybacks, what protects each business, and how far the valuations have fallen. Our view is that the market has lumped very different businesses together. RELX, LSEG and Experian sell data and decisions that an AI model cannot recreate, and look mispriced. The real risk sits where a company sells the screen rather than what's behind it: portals like Rightmove and Auto Trader, and the workflow layer at LSEG and Wolters Kluwer.
Check the claimThree claims about the AI scare, tested
A note on units. Prices are in pence (100p = £1). We convert at £1 = $1.32 and £1 = €1.18, roughly current rates. Highs and falls use weekly closing prices. P/E figures are our own calculations, dividing the share price by each company's adjusted (or "benchmark") earnings per share, which strips out one-off items and acquisition accounting.
Why are investors afraid AI will hurt RELX, LSEG and Experian?
The bear case is simple. These companies charge high, recurring fees for access to information and for the software that sits on top of it. A lawyer pays for Lexis, a trader pays for an LSEG Workspace terminal, a bank pays Experian for credit files and scores. If a general-purpose AI agent can read contracts, summarise case law or pull a company's financials, customers might need fewer licences, pay less per seat, or route around the incumbent's screen entirely.
The Anthropic plugins made that concrete. Proactive described a legal tool that could automate contract review, NDA triage, briefings and templated responses. Max Harper of Third Bridge said the fear for LSEG was that Workspace "could be rendered redundant, potentially leading to a loss of pricing power." His conclusion was more nuanced: "Pricing power may shift to data rather than platforms."
That sentence is the key to the whole sell-off. An AI model can replace an interface. It cannot replace data it has no legal or practical way to get.
The fear arrived in waves, not all at once
Last autumnRightmove's AI warningShares close down more than 12% after it guides to slower profit growth
Early this yearAnthropic's pluginsRELX down almost 17% late in the day
SpringWolters Kluwer's first quarterShares fall more than 12% despite in-line results
SpringExperian's full-year resultsRecord year; shares fall 5.1%
SummerHalf-year resultsRELX, LSEG, Sage and Pearson report faster growth
Each wave hit a different corner of the sector: portals first, then legal and data, then health and tax software.
How far have UK data and information stocks fallen?
The damage is wide but uneven. Wolters Kluwer, which sells UpToDate clinical guidance and tax and legal software, has fallen furthest. In spring its shares dropped more than 12% on a quarter that met expectations, Investing.com reported, after JPMorgan projected that UpToDate would shrink in the medium term because of customer churn and pricing pressure.
Pearson, the education group, is the outlier. It fell with the pack on the day of the Anthropic plugins but has recovered most of the way, and sits only about 9% below its high.
Wolters Kluwer has lost 60%; Pearson only 9%
Wolters Kluwer−60%
Auto Trader−48%
Rightmove−43%
Experian−38%
RELX−37%
LSEG−31%
Sage−24%
Pearson−9%
FTSE 100−4%
Latest price versus each stock's highest weekly close, excluding dividends. Most peaks came last year.
Source: Yahoo Finance weekly closes; Yield Theory calculations
What did RELX actually report?
RELX has four divisions. Risk sells fraud, identity and insurance analytics. STM publishes scientific journals through Elsevier. Legal is LexisNexis. Exhibitions runs trade shows. In the first half, group revenue was £4,871m ($6.4bn) and adjusted operating profit £1,727m, a 35.5% margin, up 0.7 percentage points.
The division the market feared most did best. Legal's underlying revenue grew 10% and its profit 13%. RELX says double-digit growth in law firms and corporate legal, about 70% of the division, is driven by Lexis+ with Protégé, its AI legal research platform with what it calls an "integrated agentic legal assistant." In other words, RELX is selling AI to lawyers and charging more for it.
Make a guess
Spot on.Close, a little low.Close, a little high.Way off, too low.Way off, too high. It was 10%.
10%, the fastest of RELX's four divisions, with underlying profit up 13%. RELX credits adoption of Lexis+ with Protégé, its AI legal platform.
Source: RELX half-year results
Risk grew 8% and STM 6%, its best in a while, helped by AI tools such as Scopus AI. Chief executive Erik Engstrom has repeated the same line all year: AI "is enabling us to add more value to our customers," and it "has been a key driver of our business for well over a decade."
Legal, the division in AI's crosshairs, grew fastest
Legal10%
Risk8%
STM6%
Exhibitions6%
Underlying revenue growth in the first half versus a year earlier.
Source: RELX half-year results
Management is also voting with the company's cash. RELX planned £2.25bn of buybacks this year, and bought £1.75bn in the first half at an average of 2,529p a share, close to today's price. Net debt rose to 2.3 times EBITDA (profit before interest, tax, depreciation and amortisation) from 2.0 times at the year-end, partly because RELX borrowed to fund those buybacks. That is still moderate leverage, but it means the buyback is no longer funded purely from cash flow.
Is LSEG's data business really at risk from AI?
LSEG is two businesses glued together: market infrastructure (the London exchange, the LCH clearing house, FTSE Russell indices) and Data & Analytics, the former Refinitiv business. The AI fear is about the second.
LSEG's half-year results were strong overall. Income excluding recoveries rose 8.4% organically, meaning without currency effects or acquisitions, to £4,799m. Adjusted EBITDA margin reached a record 52.7% and adjusted earnings per share rose 17.2% to 244.9p. LSEG raised full-year growth guidance to 7% to 7.5%.
The bears do have a point inside those numbers. Data & Analytics, the biggest division, grew only 5.1%. Within it, Workflows, which includes the Workspace terminal that competes with Bloomberg, grew just 2.8%. Data feeds grew faster.
Find the weak spot
Tap the phrase you think is the tell.
Spotted it.Not that phrase. Workflows grew just 2.8%. If AI agents become the way analysts pull data, the terminal matters less and the data matters more.
“Markets: +11.9%” Trading and clearing boomed. AI can't clear a swap.
“FTSE Russell: +9.1%” Index licences are proprietary standards. Strong growth.
“Data & Analytics: +5.1%” Slower, but data feeds within it grew well. Look one level down.
“Workflows, including Workspace: +2.8%” This is the screen layer, the part an AI agent could most easily replace.
Source: LSEG half-year results
LSEG's answer is to sell its data to the AI companies rather than fight them. It credits three years of work with Microsoft on its data estate, and now integrates Workspace with Microsoft Copilot. It has added partnerships with Amazon and Google, puts its data inside Anthropic's Claude, and makes data available to AI models through MCP, the Model Context Protocol, an open standard that lets AI assistants connect to outside data sources. LSEG says more than 200 customers are using its AI-ready data this way.
In its full-year results presentation, titled "More valuable in an AI world", LSEG claimed about 98% of group revenue comes from proprietary data, intellectual property and market infrastructure. Only about 2% is non-proprietary data such as economic statistics and as-reported financials.
Watch it play outOnly 2 pounds in 100 of LSEG's revenue is data anyone could find
Picture every £100 LSEG earns as 100 dots.
LSEG says about £98 comes from proprietary data, intellectual property and market infrastructure.
Only about £2 is non-proprietary data, like economic statistics, that rarely sells on its own.
Its plan: pipe that data into AI tools from Microsoft, Amazon, Google and Anthropic, and get paid either way.
Source: LSEG full-year results presentation
That claim is generous, since it counts the whole of Data & Analytics, about 42% of revenue, as proprietary, including a terminal that faces real pressure. But the rest, the exchange, clearing and FTSE Russell, is hard to disrupt with a chatbot. You can't clear a trade or set an index benchmark with a language model.
How exposed is Experian?
Experian keeps credit files on people and companies and sells scores and decision tools to lenders. It also runs a consumer business with more than 215 million free members. Its full-year results were the best in its history: organic revenue growth of 8%, a benchmark operating margin of 28.6%, and benchmark earnings per share up 15% to 179.8 US cents. It added a fresh $1bn buyback. The shares fell 5.1% on the day, according to AJ Bell.
Its latest quarterly update showed 7% organic growth, with business-to-business revenue up 9% and consumer services just 2%. Management says full-year expectations of 6% to 8% organic growth are unchanged.
We think Experian's credit data is among the hardest assets in this group for AI to copy. It is collected under regulation from lenders who report repayment histories, and lenders rely on it for regulated decisions. An AI model can help a lender use that data; it can't make it. The softer spot is the consumer arm, which competes for attention with free apps and AI assistants.
What about Wolters Kluwer, Sage and Pearson?
Wolters Kluwer is the cautionary tale. Its half-year report shows organic growth of 5% and an adjusted operating margin of 29.4%, up a point. UpToDate Expert AI has been adopted by more than 90% of its US enterprise customers. Yet its shares are down 60%, because UpToDate, a doctor's reference tool, is closer to the "answer engine" AI models are learning to be.
Sage, the accounting software group, is down 24% and has bounced off its lows. Its nine-month update showed underlying revenue growth accelerating to 11%, with Sage Copilot and other AI features available to more than 600,000 customers. A ledger that small businesses trust with their books and tax returns is a workflow, not a search box.
Pearson has turned AI into a sales pitch. Its interim results showed 4% underlying sales growth and adjusted operating profit up 14%, and it signed a deal to run a leading AI lab's global certification programme. The market has largely accepted that pitch.
Why are Rightmove and Auto Trader different?
The portals sell something else: an audience. Rightmove is Britain's biggest property portal, and estate agents pay because that's where buyers look. Auto Trader plays the same role for used cars. Both run operating margins near 70%, a sign of enormous pricing power.
The AI risk is that buyers start their search in a chatbot instead. If an AI assistant can find the three-bedroom house or the used hatchback and send the buyer straight to the agent or dealer, the portal's audience, and with it the moat, shrinks. That is a different, more direct threat than the one facing RELX or Experian.
Rightmove's own response spooked investors first. Last autumn it said it would spend more on AI, guiding to just 3% to 5% growth in underlying operating profit this year, Reuters reported. The shares closed more than 12% lower. Its half-year results showed revenue up 7% and a margin of 69%, but it cut its revenue growth guidance to 6% to 8% because fewer new-build developments were advertised.
Auto Trader's problem is partly self-inflicted. Its full-year results showed revenue up just 4% to £624.3m after dealers rebelled against the rollout of Deal Builder, its online car-buying tool. Growth slowed to 3% in the second half and, according to Yahoo Finance, revenue was flat in the first month of the new year.
The moat that AI can't copy versus the one it can route around
Hard for AI to replace
Credit files reported by lenders (Experian)
Exchange, clearing and indices (LSEG)
Fraud and identity data (RELX Risk)
Journal archives and citations (RELX STM)
Accounting ledgers small firms trust (Sage)
vs
Easier to route around
Portal audiences (Rightmove, Auto Trader)
Terminal screens (LSEG Workspace)
Reference answers (Wolters Kluwer UpToDate)
Language-learning products (Pearson)
How much cheaper are UK data stocks now?
The de-rating is large. At their peaks, RELX, LSEG and Experian all traded at roughly 33 to 35 times the latest full year's earnings, typical of high-quality compounders. Now all three trade around 18 to 19.5 times.
Company
Peak P/E
P/E now
EPS used now
RELX
34x
19.5x
133.6p, last twelve months
LSEG
33x
18x
456.6p, last twelve months
Experian
35x
19x
179.8 US cents, last full year
Auto Trader
29x
14x
34.17p basic, last full year
For LSEG, trailing earnings of 456.6p combine last year's 420.6p with the latest half-year. Auto Trader's peak multiple is approximate, because we estimated the prior year's earnings from its reported 8% growth.
Investors now pay about half the multiple they paid at the peak
P/E at peakP/E now
10203040x0
34.3x19.5x
RELXP/E at peak34.3xP/E now19.5x
33.1x18.2x
LSEGP/E at peak33.1xP/E now18.2x
35.0x18.7x
ExperianP/E at peak35.0xP/E now18.7x
28.7x13.8x
Auto TraderP/E at peak28.7xP/E now13.8x
RELXLSEGExperianAuto Trader
Price divided by adjusted or benchmark earnings per share; Auto Trader uses basic EPS. Peak uses the latest full year reported at the time.Source: Company results; Yahoo Finance; Yield Theory calculations
At 19.5 times, the market is not saying RELX will collapse. It's saying RELX's growth won't last long enough to deserve a premium. You can test that. Try different rates of earnings growth over three years and the P/E investors pay at the end.
RELX only loses money if growth slows and the multiple stays crushed
Pick RELX's yearly EPS growth for three years and the P/E at the end
Share price in three years2,190p2,847p3,650p2,455p3,191p4,092p2,741p3,563p4,568p
Starts from 133.6p of trailing adjusted EPS. Excludes dividends, which add roughly 2.6% a year at the current 67.5p payout.Source: Yield Theory calculations
Which UK data stocks are mispriced, and which face real disruption?
Our ranking, from most mispriced to most at risk:
Experian. Regulated, lender-fed data, double-digit earnings growth guided again, and a P/E below 19. AI is more likely to raise demand for its data through better lending decisions than to replace it.
RELX. Its fastest-growing division is the one AI was supposed to kill. Risk and STM depend on proprietary data and archives. The buyback at around 2,529p a share looks well timed.
LSEG. Most of the group, exchanges, clearing and indices, is untouched by AI. The Workspace terminal is a real weak point, which is why we rank it third.
Sage and Pearson. Both are winning so far, and their shares have recovered. Less mispriced, less at risk.
Wolters Kluwer. Growing, but its clinical answers business sits closest to what general AI does well. Cheap for a reason, partly.
Rightmove and Auto Trader. The most genuine disruption risk. Their moat is where buyers start a search, and that habit is exactly what AI assistants are trying to change. Auto Trader's low P/E reflects its own execution stumbles as well.
What could go wrong with the bull case?
Seat-based pricing could erode slowly before it shows in revenue: multi-year contracts delay the damage. Customers might build their own tools on cheaper data. RELX's borrowing to buy back shares raises the stakes if growth slows. Data & Analytics is already LSEG's slowest-growing division. And a lower P/E isn't a floor; Auto Trader shows multiples can keep falling when growth disappoints.
Our read
The AI sell-off in UK information stocks treated a property portal, a stock exchange and a credit bureau as the same bet. They aren't. The question that separates them is whether an AI agent needs the company's data or just its interface.
RELX, LSEG and Experian are still growing revenue at 7% to 8%, expanding margins, and buying back stock at prices far below their peaks. Their P/E multiples have roughly halved. For RELX and Experian especially, we think the market is pricing a disruption that hasn't appeared in results and that their data makes unlikely.
LSEG is cheaper than it looks if you value the exchange, clearing and index businesses on their own. The terminal is the part to watch, and its slow growth is the honest bear argument.
The portals are where we'd be most careful. Rightmove and Auto Trader own audiences, not irreplaceable data, and their low multiples may still not fully price a world where buyers search through an AI assistant.
Go deeper
Competitive advantage: how to tell a data moat from an audience moat, the line that splits this sector.
RELX trades 37% below its peak and at about 19.5 times earnings, down from 34, even though its profit per share keeps rising. We test the AI fears against what RELX, LSEG, Experian, Wolters Kluwer, Sage, Pearson, Rightmove and Auto Trader actually reported.
Unilever spun off €7.9 billion of ice cream and is handing €10.7 billion of food to McCormick. What the slimmer company earns, how Magnum Ice Cream has traded, and whether either stock deserves a higher price.
The thesis, the numbers behind it, and what would break it. Full access is $39 a month.
FTSE 100, London's index of its 100 largest listed companies, sits just 4% below its own record.