GoogleNovember2025AI

and

theFutureofLearningBen

GomesChief

TechnologistLearning&SustainabilityChristopherPhillipsVP&

GMEducationLila

IbrahimChiefOperatingOfficerGoogleDeepMindJamesManyikaSVPResearch,Labs,

Technology&SocietyYossiMatiasVP&

GMGoogleResearchChallengesImportantAreastoAddressOpportunitiesAI’s

Promisefor

LearningOurWorkonAIin

LearningLearning

MattersConclusionAbstract040506030201Googlebelievesthatbysupportingteachersandpersonalizinglearning,

AI

can

help

unlock

human

potential

all

around

the

world.While

human

relationships

will

always

remain

central

to

teachingand

learning,we

see

tremendous

opportunity

for

AI

to

supportGoogle’s

mission

to

make

the

world’s

information

universallyaccessible

and

useful.

At

the

same

time,

AI

presents

urgentchallengesandunknownsthatsocietymustcollectivelyreckonwith.Our

goal

is

to

help

improve

learning

outcomes

by

developingAI

products

that

are

grounded

in

core

learning

science

and

inclose

partnership

with

the

education

community—because

thegreatest

potential

of

AI

is

helping

everyone

reach

theirs.Google

AI

and

the

Future

of

Learning

3AbstractAIopensupnewpossibilitiesfor

people

tolearn,accesseducation,andfollow

theircuriosity.

Itcanalsohelpaddress

criticalchallenges,includingconcerning

trendsinlearningoutcomes,immense

pressureoneducators,andalackofaccesstoqualityeducationinmany

communities.LearningMattersLearning

is

the

bedrock

of

human

potential

and

societal

progress.Itequips

us

with

new

skills,sparks

curiosity

and

the

exchange

of

freshideas,spurs

innovative

breakthroughs,and

drives

economic

growth.New

technologies

have

always

been

a

force

to

expand

society’sknowledge,as

well

as

accelerate,challenge,and

ultimately

reshape

how

welearnandwhatweknow.Fromtheinventionoftheprintingpresstothe

vast

expanse

of

the

internet,advances

in

technology

have

helpedto

democratize

access

to

information

and

opportunity,pushing

theboundaries

of

human

potential.Learning,while

lifelong,is

especiallyrapid

in

early

childhood—both

within

and

outside

of

formal

educationsettings.Curiosity

and

the

desire

to

learn

are

human

nature,

and

today,learning

opportunities

are

accessible

to

more

people

than

ever

before.Google

AI

and

the

Future

of

Learning

4<1%1-5%

5-10%10-30%>30%Bucketed

by

resources

available

per

young

person,relative

to

USA????}????Most

of

the

world

has

lessthan10%ofUSAresources(GDP)per

youngperson

with

whichtoeducatetheirpopulation.Historically,formal

education

was

exclusive:limitingthe

path

to

opportunities

to

a

select

few.

As

a

result,many

efforts

have

focused

on

widening

access,witha

long

arc

of

progress

from

royal

tutors(effective,personalized,but

inaccessible),to

the

industrialmodel

of

education(a

powerful

tool

for

expandingaccess

and

literacy,but

not

personalized),to

recent

educationtechnologiesthathaveallowedteachersto

reachlearnersbeyondtheirclassrooms.The

world

has

made

huge

strides

in

access.

90%of

primaryschool-agedchildrengloballyareenrolled

in

school

and

approximately

87%complete

theirprimary

education.However,challenges

remain,including

uneven

access

to

education,differences

in

education

quality,limited

content

availability

across

languages,as

well

as

low

and

unequal

learningoutcomes.Abriefhistoryofeducation:Wherewestarted,

where

weareGoogle

AI

and

the

Future

of

Learning

5Global

population

of

young

peopleaged0-19Resources

defined

as2023GDP

at

PPP,

from

World

Bank;

young

person

defined

as

aged0-19,

from

UN.Grey

circles

have

no

GDP

data

available.Per

young

person

resources

relative

to

USALearning

mattersMaking

good

on

the

promise

of

education

meansensuringalllearnersreachproficiencywhilesupporting

their

potential

and

well-being

at

everylevel.Publiceducationsystems

facesignificantstrain

as

a

result

of

inadequate

funding,thecontinuingimpactof

pandemic-eralearningloss,

disruption

fromglobalconflicts,teacherturnoverand

shortages,student

attendance

and

mentalhealthchallenges,andpersistentsocio-economic

disparities.Theseinterconnected

factorsarecompoundedbythe

need

to

constantly

update

curricula

to

preparestudents

for

the

demands

of

a

changing

world.AccordingtotheOECD’sProgramme

forInternational

Student

Assessment(PISA),globallearning

outcomes

have

declined

over

the

last

twodecades.Results

from

the

2022PISA

survey,whichcaptured

data

from81countries

and

economies,notedan“unprecedentedperformancedrop”

withmean

performance

in

mathematics

down

by15points

and

reading

down

by10points

compared

to2018.Challenges

are

being

seen

in

developed

anddevelopingcountriesalike.

Anddespitepositivetrendsinenrollment,

272millionchildrenremainoutofschool(11%ofchildreninprimary,15%inlowersecondary,31%inuppersecondary).

Astheschool

age

population

of

low-and

middle-incomecommunities

continues

to

grow,experts

estimate

the

world

will

need

44million

more

teachers

by2030

toprovideuniversalprimaryandsecondaryeducation,

includinginplacesthatalready

facemeaningfulbarriers

to

fully

supporting

their

existing

students.It

is

not

easy

to

keep

up

with

the

acceleratingdemands

of

a

rapidly

changing

world.

Today’s

jobsdemand

not

just

foundational

skills,but

advancedproblem

solving,collaboration,and

the

ability

tolearn

throughout

life.

The

emergence

of

AI

marksanother

period

of

profound

transformation—theopportunity

to

rethink

how

education

should

adapt

to

both

the

opportunities

and

the

challenges

ahead.

Aseducationalsystemsgrapple

withthequestionofwhat

it

means

to

be

well

educated

in

a

world

where

AIis

ubiquitous,students

will

need

a

new

set

of

skills

toprepare

forthe

future.Trends

in

mathematics,

reading

and

scienceperformancePISA

test

scores,OECD

average MathReading

ScienceLearning,teachingandeducationare

universal

in

humansocietiesLearning:The

acquisition

and

retention

of

information,representingashiftinwhatanindividualknowsorcan

do.Teaching:The

process

of

helping

someone

else

learn.Thishappens

in

both

formal

education

and

outside

of

that

process.Education:The

bundling

of

teaching

and

learning

in

a

structuredsetting

alongside

a

relatively

consistent

curriculum.

This

oftenends

with

a

formal

recognition

of

completion.For

simplicity,we

use

the

term

teacher

to

indicate

an

educator

in

this

type

offormal

setting.Though

focusedonprimaryandsecondaryeducation,this

paper'sdiscussionofopportunities,limitations,and

futureexploration

is

relevant

to

all

ages.Google

AI

and

the

Future

of

Learning

6Note:

AIisabroadtermencompassinga

widerangeof

technologies,

includingpredictivemodels,generative

AIrootedinlargelanguage

models,andgeneralintelligence.

Throughoutthispaper,

weusethe

term

AItotalkabout

the

category

as

a

whole.Source:

OECD(2023),PISA2022Results(VolumeI):

TheStateof

Learning

and

Equity

in

Education.Learning

mattersOpportunities:AI’sPromise

forLearningAdvances

in

AI,such

as

large

language

models,present

newopportunitiestoequiplearners

withtheskillstheyneed

forsuccess.While

the

internet

removed

barriers

to

accessing

information,

AIincreases

our

ability

to

understand

and

apply

that

information.

This

leapfrom

passive

consumption

to

active,deep

understanding

is

a

profoundchange

that

is

revolutionary

for

education.

AI

can

expand

teaching

andlearning

by

modifying

content

to

be

more

engaging

and

interactive,simplifying

complex

topics,and

enabling

deeply

personalized

learning

atscale.Google

AI

and

the

Future

of

Learning

7Education

systems

have

always

faceda

core

challenge:limited

resources.Ideally,everystudentwouldspendampletimeworking

in

their

zone

of

proximal

development–thesweet

spot

of“just

right”learning

challenges

thatlead

to

new

skill

growth.But

because

student

needs

vary

so

widely,traditional

education

cannot

alwaysprovidethispersonalizedsupporttoeveryone.While

AI

is

by

no

means

perfect,it

does

have

thepotential

to

reduce

barriers

and

allow

people

to

learn

more

effectively

than

before.For

students,

AI

canhelp

create

individualized

learning

paths

and

offerextra

help,whether

or

not

they

have

the

supportof

an

educator.For

educators,

AI

can

serve

as

ateachingassistantthatsupportstheirworkload

andenables

new

approaches,ultimately

freeing

up

moretime

for

the

essential

human

aspects

of

teaching.It

iscriticalthat

AI-poweredlearningtoolsaredevelopedand

supervised

by

experts

and

educators

to

ensure

they

reflect

learners’needs

and

provide

effective,evenlydistributedpedagogicalbenefits.Technologists

have

made

similar

promises

aboutaccessandpersonalizationduringprior

wavesof

education

technology—and

while

some

weremet,many

were

not.

These

cycles

did,however,offer

valuable

lessons

about

what

works(andwhat

doesn’t)in

education

technology,includingthattechnologyalonecannotsolvedeep-rootedchallenges,that

it

must

be

designed

with

and

for

thestudents

and

teachers

it

intends

to

reach,and

thatusers

must

be

equipped

with

the

knowledge

around

howtoimplementnewtoolsinmeaningfulways.AI

is

far

more

powerful

than

past

innovations,offering

truly

transformative

possibilities.However,turningAI’s

promise

into

reliable,everyday

practice

willrequire

both

time

and

intentional

effort."WhileAIisbynomeansperfect,

itdoes

have

the

potentialtoreducebarriersandallowpeopleto

learnmoreeffectivelythanbefore."Google

AI

and

the

Future

of

Learning

8Opportunities:AI’s

Promisefor

LearningWhendesignedandimplementedcorrectly,AI

systems

can

embody

the

proven

principlesof

learning

science:including

sparking

activeparticipation,encouraging

deep

practice

and

review,

and

weaving

in

the

benefits

of

spaced

repetition.

Thiscanenhancelearningeffectivenessandefficiency,leading

to

more

breakthrough

moments

for

morestudents.

AI

systems

that

are

explicitly

fine

tuned

forthese

purposes

can

create

new,scalable

learningactivitiesandassessmentsthatwerepreviouslyimpossible.For

educators,these

tools

can

lightenthe

burden

of

implementation,making

powerfulteachingstrategiesmoreaccessibleandachievablein

less

time.For

students

who

choose

to

engage,AI

moves

them

from

passive

consumption

to

activeapplication—becoming

a

catalyst

for

curiosity,moremeaningful

learning,and

assessments

that

feel

lesslike

memorization

tests

and

more

like

opportunities

todemonstrateknowledge.The

potential

goes

even

further:

AI

can

help

usdeepen

our

understanding

of

learning

itself.By

analyzing

patterns

of

student

engagement,motivation,and“aha”moments,we

can

uncover

new

insights

into

how

curiosity

grows,how

persistencetakes

root,and

how

learning

flourishes

acrossdifferentcontexts.Perhaps

learningalongside

AImayreveal

entirely

new

facets

of

skill

development

andhuman

growth

potential—reminding

us

that

learning

is

not

just

about

knowledge

transfer,but

aboutexpanding

the

boundaries

of

what

is

possible.interactive_spacebook_ribbondoor_opencheerPersonalizingMake(almost)RemovingAhelpingteachingandtutoringanythinglearnablebarriers

tolearninghand

forbusyeducators01UnlockingthepoweroflearningscienceAIpromisestobringthe

verybestof

what

weknowabouthowpeoplelearn(learningscience)intoeverydayteaching,andmayeven

identify

new

frontiers

for

education

research.Google

AI

and

the

Future

of

Learning

9HowAIcantransformlearninglab_researchUnlockingthepoweroflearningscienceOpportunities:AI’s

Promisefor

LearningOutsidetheclassroom,lackof

foundationalknowledge,study

skills,motivation,or

means

canall

present

barriers

for

studying

or

completinghomework;as

can

simply

finding

someone

tohelp

you

when

you’re

stuck.

AI

can

serve

as

aninexpensive,non-judgemental,always-availabletutor.

There

is

rigorous

evidence

that“high-dosage”personal

human

tutoring

has

one

of

the

largestpositive

impacts

on

student

achievement,withconsistentlysignificanteffectsonstudents

fromfamilies

with

low

incomes1,2,3,4

.

While

AI

tutors

cannotreplace

the

human

element

of

great

teaching

ortutoring,they

can

be

a

complement

or

bridge

when

human

support

is

unavailable.

As

AI

tutors

becomemore

widely

used,it

will

be

important

to

understandhow

much

of

the

benefit

from

tutoring

comes

fromthe

opportunity

to

work

within

the

proximal

zone

ofdevelopment

with

a

knowledgeable(potentially

AI)coach

versus

the

power

of

human

connection(suchas

the

natural

desire

to

connect

with

someone

whocares

about

you).Whilestructurededucationhashadsuccessinteaching

large

groups,it

often

struggles

to

meetthe

diverse

needs

of

individual

students.

AI

mayultimately

allow

every

learner

to

take

a

trulyindividualized

learning

journey,realizing

the

promiseof

personalization

like

never

before.AI

can

give

educators

tools

that

can

adapt

lessoncontent

to

a

student’s

specific

needs

and

goals—providing

tailored

feedback

and

identifyingknowledge

gaps

that

hinder

layered

learningprogress.It

can

also

adjust

based

on

a

student’scontext,like

whether

a

teacher

is

present,whatthe

subject

is,and

what

has

been

learned

before.Educators

can

receive

insights

into

student

progress

and

learning

gaps,enabling

more

personalized

andfocusedsupport

whilereducingcognitiveload.

Thiscan

help

target

learners

at

the

right

level

to

close

understanding

gaps

and

help

ensure

that

learningdifferences(including

neurodivergence)do

notdetermineone'spotential

forsuccess.interactive_spacebook_ribbondoor_opencheerPersonalizingMake(almost)RemovingAhelpingteachingandtutoringanythinglearnablebarriers

tolearninghand

forbusyeducatorsThegoalofpersonalizationisn’tnew,but

AIopensupnew

waystobringittolifeatscale.

What

wasonceanaspirationmaynowbe

possible.02PersonalizingteachingandtutoringGoogle

AI

and

the

Future

of

Learning

10lab_researchUnlockingthepoweroflearningscienceOpportunities:AI’s

Promisefor

LearningPeople

are

most

curious

about

things

that

theyfeel

are

learnable,or

that

they

are

on

the

vergeof

understanding.

AI

can

strip

away

jargon

andadjust

explanations

and

delivery

modes

to

alearner’s

background

and

preferences—helpingstudents

spend

more

time

in

their

zone

of

proximaldevelopment.Thecoreschoolcurriculumiswell

covered

withexpertly

produced

content,but

quality

options

fall

offrapidly

when

moving

into

more

specialized

subjects.AI

opens

up

new

opportunities

for

people

to

learnand

retrain

throughout

their

lives,especially

for

nicheinterests

and

changing

job

market

demands.Whetherit’sconnectingacademicconceptstoreal-

world

applications

or

bridging

the

gap

between

newinformation

and

prior

knowledge,

AI

tools

can

helpmake

learning

more

exciting–

fostering

intrinsicmotivation

and

sparking

curiosity

and

exploration.interactive_spacebook_ribbondoor_opencheerPersonalizingMake(almost)RemovingAhelpingteachingandtutoringanythinglearnablebarriers

tolearninghand

forbusyeducatorsWeliveinanageofinformationabundance.Inorder

tobeunderstandable,informationneedstobeavailableand

approachable.03Make(almost)anythinglearnableGoogle

AI

and

the

Future

of

Learning

11Fromlefttoright,toptobottom:ExploretheMuséed’Orsay,Paris

with

GoogleAI,

Woolaroo,

Google

Lens,

Moving

paintings.lab_researchUnlockingthepoweroflearningscienceOpportunities:AI’s

Promisefor

Learninginteractive_spacebook_ribbondoor_opencheerPersonalizingMake(almost)RemovingAhelpingteachingandtutoringanythinglearnablebarriers

tolearninghand

forbusyeducatorsFor

the

majority

of

human

history,access

toknowledge

and

information

was

a

critical

barrier

tolearning.

While

the

printing

press

and

internet

haveimproved

access

in

recent

history,information

isstill

not

universally

available.Some

of

this

remainsan

infrastructure

challenge,but

barriers

exist

evenfor

those

with

internet

access:a

scarcity

of

contentand

learning

opportunities

in

one’s

native

language,

content

format

limitations,and

content

unsuitablefor

different

learning

levels.

AI

has

the

power

to

helpin

all

of

these

respects

by

transforming

content

intonew,more

accessible

forms,and

delivering

contentat

the

right

level

to

close

learning

gaps.Millions

of

learners

still

lack

access

to

formaleducation

or

sufficient

time

with

a

teacher,or

facechallenges

fromphysicalorlearningdisabilities.AI

has

the

potential

to

deliver

quality

learningexperiences

—making

the

world’s

information

more

suitable

to

individuals’needs.AImaynotonlyraisetheceilingforthosewhohave

educational

support

today,it

will

also

significantly

raisethe

floor

forlearnersacrosstheglobe.Everyone

deserves

access

to

quality

learning

experiences

andtheskillsandtoolsneededtobuild

the

future

theyenvision

forthemselves.04RemovingbarrierstolearningGoogleAI

and

the

Future

of

Learning

12lab_researchUnlockingthepoweroflearningscienceOpportunities:AI’s

Promisefor

Learninginteractive_spacebook_ribbondoor_opencheerPersonalizingMake(almost)RemovingAhelpingteachingandtutoringanythinglearnablebarriers

tolearninghand

forbusyeducatorsAI

can

assist

busy

teachers

across

a

range

of

areas,

spanning

fromdeepresearchandcontentcreation,to

activity

planning

and

administrative

tasks.Separately,

AI

can

also

assist

by

enabling

new

formsof

instruction

or

assessment.

This

can

include

AI-ledgroup

breakouts

that

drive

active

student

learning,

adaptivereal-time

feedback

formultiplestudentsat

once,or

open-ended

debates

that

evaluatearguments

and

push

the

frontier

of

a

student’s

understanding.This

assistance

can

both

help

students

fill

gaps

intheir

learning

and

enable

teachers

to

spend

more

time

with

students.Most

importantly,

AI

can

helpteachers

get

back

to

doing

the

job

they

signed

upfor,

focusing

on

the

important

human

aspects

of

teaching:helping

their

students

learn

how

to

learn,

inspiringtheirinterests,

fosteringconnections,and

developing

them

as

curious

and

well-roundedindividuals.AI

can

become

a

powerful

teaching

assistant,

freeing

up

precioustime

for

teachers

to

focus

on

meaningful

human

and

socialinteraction

with

students.05AhelpinghandforbusyeducatorsGoogleAI

and

the

Future

of

Learning

13lab_researchUnlockingthepoweroflearningscienceOpportunities:AI’s

Promisefor

LearningChallenges:Important

Areas

to

AddressLike

all

major

technological

shifts,

AI

introduces

new

challenges.Someare

familiar,such

as

protecting

privacy

and

preventing

exposure

toinappropriate

content,while

others

are

specific

to

AI’s

capabilities.

Whilethesechallenges

varytremendouslyacrossglobalsettings,andmanifestdifferently

across

primary,secondary

and

higher

education

levels,they

must

be

confronted

head-on.Questions

of

this

nature-and

magnitude-

are

best

answered

together,and

help

us

make

progress

as

a

society.Google

AI

and

the

Future

of

Learning

14While

hallucination

rates

have

fallen

sharply

in

thepast

two

years

as

models

are

trained

to

use

trustedsources

and

verify

the

outputs,a

harder

challengeremains:determining

which

sources

are

trustworthyand

how

to

handle

subjective

questions.

This

isparticularly

challenging

given

the

biases

in

globaldata.AI

also

needs

the

right

tone

and

must

avoidconfirmingincorrectassertionsandprovidingexcessive

praise.It

should

challenge

a

student’smisconceptionsandcorrectinaccuratestatements,not

simply

act

as

an

uncritical

mirror

that

reinforcesthem.

This

is

important

both

in

ensuring

accuracy,but

also

in

maintaining

educational

motivation

and

self-regulation.This

is

an

area

of

deep

importance

to

Google,

andwhile

no

system

is

perfect,we

have

made

meaningfulprogress

with

each

new

model

release,and

continue

driving

the

industry

forward

with

new

benchmarks.

Bycarefullyandresponsiblybuildinganddeploying

AImodels,weaimtomanagetheserisksaswehave

for

our

other

businesses,like

Search,since

ourearliest

days.targetshieldnetwork_intelligencepsychology_altcrowdsourceAccuracyandobjectivitySafetyCriticalthinkingCheatingandlearninglossEqualaccess01AccuracyandobjectivityA

primary

concern

is

that

AI

models

can"hallucinate"and

producefalseormisleadinginformation,similarto

humanconfabulation.Google

AI

and

the

Future

of

Learning

15Challenges:

ImportantAreastoAddressA

layered

safety

strategy,including

age-appropriate

AI

literacy

and

content

filters

is

essential,but

willneed

to

evolve

as

fast

as

the

technology

does.Google’s

commitment

to

safety

is

underpinned

byextensive

testing

and

continuous

expert

consultation.

We

conduct

rigorous

user

and

adversarial

testing,including“Red

Teaming”exercises

and

child

safety-

focusedsprints,specificallydesignedtoidentifyvulnerabilitiesandbuildstrongsafeguards