Our forecast, stated as a forecast: every advertising platform we can measure let its early users in cheap while the audience was being built, then raised the price for everyone who came later. We expect AI recommendation to follow the same path. The difference this time is the prize: not one of ten links, but one of the two to four names the AI says out loud.
This page is the evidence behind that forecast. Every number comes from a company’s own filing or a named study, listed under Sources. Where we are guessing, we say so in the sentence.
Google Ads, then versus now
Then. In 2001 Google’s whole business was $86.4 million a year.[1] On February 20, 2002 it opened a self-serve auction in which advertisers “pay only when their ads are clicked on by users.”[2] Google described cost-per-click as a model it had “launched in 2002.”[1] The auction was open to anyone, and the whole of it was the size of a mid-sized company.
Now. In 2025 Google sold $294.7 billion of advertising, about 3,400 times its 2001 revenue.[1][3] A US small business pays an average of $5.42 every time someone clicks a search ad, measured across thousands of LocaliQ and WordStream customer campaigns.[5] And Google’s own average price per click, which fell for most of the 2010s as it added billions of cheap mobile clicks, has risen in four of the last five years: +15% in 2021, +1% in 2023, +7% in 2024 and +7% in 2025.[4][3]
One caveat, because it matters. Google’s disclosed average is global and mix-weighted. When cheap clicks (mobile, emerging markets) grow faster than expensive ones, the average falls even while a local plumbing click gets dearer, and Google says exactly that in its own filings.[4] So the chart below is not “what a plumber pays.” It is Google’s own price index, and the direction since 2021 is up.
Google's own average price per click, indexed
Chained from each year's change in average cost-per-click as disclosed in Google's and Alphabet's 10-K filings. 2007 = 100. Global and mix-weighted: it fell as billions of cheap mobile clicks arrived, and it has risen since 2021.
| Year | Index (2007 = 100) | Reported change |
|---|---|---|
| 2008 | 107 | +7% |
| 2009 | 99.5 | −7% |
| 2010 | 104.5 | +5% |
| 2011 | 107.6 | +3% |
| 2012 | 94.7 | −12% |
| 2013 | 87.1 | −8% |
| 2014 | 82.8 | −5% |
| 2015 | 73.7 | −11% |
| 2016 | 65.6 | −11% |
| 2017 | 53.1 | −19% |
| 2018 | 39.8 | −25% |
| 2019 | 37.4 | −6% |
| 2020 | 33.7 | −10% |
| 2021 | 38.8 | +15% |
| 2022 | 38.4 | −1% |
| 2023 | 38.7 | +1% |
| 2024 | 41.5 | +7% |
| 2025 | 44.4 | +7% |
The index chains each year’s disclosed change with 2007 set to 100. Since 2020 it is up about 32%.[4]
Facebook and Meta, then versus now
Then. Facebook Pages launched on November 6, 2007. They were free, Insights was “a free service for all Facebook Pages,” and more than 100,000 Pages were created on day one.[6] In 2012 a Page post reached about 16% of the people who had liked the Page, for free.[7]
Now. By early 2014 that free reach was 6.5%, and about 2% for Pages with 500,000 or more likes.[7] Facebook explained why in 2014: there were about 1,500 stories competing for each person’s feed and only about 300 were shown, so “it’s becoming harder for any story to gain exposure.”[8] The free channel narrowed and the paid one grew. Meta’s own average price per ad rose +24% in 2010 and +18% in 2011,[9] +173% in 2014, +140% in 2015, +29% in 2017 and +24% in 2021,[10] +10% in 2024 and +9% in 2025,[11] and +12% in the most recent quarter.[12] Meta sold $26.9 billion of ads in 2016 and $196.2 billion in 2025.[13][11]
Caveat, again. The 2014 and 2015 jumps include a product change: Facebook showed fewer ads and made each one more prominent, so the price per ad unit leapt.[10] It is still what advertisers paid per ad. Starting the clock in 2016, after the format change, the index is up about 42%.[10]
Meta's average price per ad, indexed
Chained from each year's change in average price per ad as disclosed in Facebook's S-1 and Meta's 10-K filings and earnings releases. 2009 = 100. The 2014 and 2015 jumps include a format change: fewer, more prominent ads.
| Year | Index (2009 = 100) | Reported change |
|---|---|---|
| 2009 | 100 | base |
| 2010 | 124 | +24% |
| 2011 | 146 | +18% |
| 2012 | 151 | +3% |
| 2013 | 205 | +36% |
| 2014 | 560 | +173%, format change |
| 2015 | 1,343 | +140%, format change |
| 2016 | 1,410 | +5% |
| 2017 | 1,819 | +29% |
| 2018 | 2,056 | +13% |
| 2019 | 1,953 | −5% |
| 2020 | 1,757 | −10% |
| 2021 | 2,179 | +24% |
| 2022 | 1,831 | −16% |
| 2023 | 1,666 | −9% |
| 2024 | 1,832 | +10% |
| 2025 | 1,997 | +9% |
The same pattern everywhere
Amazon. Amazon’s advertising business went from about $1.7 billion in 2015, reported as “Other,” to $68.6 billion in 2025.[15][16] A marketplace analyst estimates that Amazon now keeps more than half of what a seller makes, across referral fees, fulfillment and advertising, up from about 40% five years earlier, because “most of the best-converting screen space is allocated to advertising.”[14] That is the analyst’s estimate, not an Amazon figure.
Uber and DoorDash. Uber kept about 20 cents of every booking dollar in 2016 and about 27 cents in 2025.[17][18] DoorDash’s revenue as a share of marketplace order value went from 11.0% in 2019 to 13.4% in 2025.[19][20] Both ratios are directional. The accounting changed over the period.
Yelp. Yelp let businesses claim a listing for free,[23] then grew from about 7,000 paying local business accounts at the end of 2009[21] to about 565,000 paying advertising locations at the end of 2019.[22]
The shape is the same each time. Build the audience with a free or cheap channel, then charge for the position in front of it.
What is different this time
Search gave every business a chance at ten links. AI gives a few businesses a name in a sentence.
In a study of 200,085 local searches, ChatGPT named an average of 4.1 businesses per answer, Google AI Mode 3.5 and AI Overviews 2.5.[24] A tracked business appeared in about a third of AI answers (33% in ChatGPT, 32% in AI Mode, 38% in AI Overviews), against 66% of Google Maps results.[24] And the number of people asking moved in a single year: 45% of US consumers used an AI tool for a local recommendation in 2026, up from 6% in 2025.[25]
So the shortlist is short, the odds of being on it are about one in three, and the audience is arriving now. The companies that built Google and Facebook know what a business will pay to be the one that gets picked. They sold $295 billion and $196 billion of ads last year.[3][11]
Pay-to-play is already dated
Paid placement next to AI answers is not a prediction. It has dates.
| Date | Platform | What happened |
|---|---|---|
| March 29, 2023 | Microsoft Bing chat | Said it was “exploring placing ads in the chat experience.”[26] |
| October 3, 2024 | Google AI Overviews | Ads in AI Overviews for US mobile users, labelled “Sponsored.”[27] |
| November 12, 2024 | Perplexity | “Sponsored follow-up questions,” with launch brands including Indeed and Whole Foods.[28] |
| May 21, 2025 | Google AI Mode | “Expanding ads in AI Overviews to desktop, and bringing ads to AI Mode.”[29] |
| January 16, 2026 | OpenAI, ChatGPT | Announced ads for logged-in adult US users on the free tier.[30] |
| September 1, 2026 | OpenAI, ChatGPT | “$1 billion in annualized revenue run rate” in under 200 days, with geographic targeting and outcome-optimized bidding in the Ads Manager.[31] |
Two things to be clear about. First, every one of those is an ad placed beside or under an answer, labelled as an ad. We have not found any platform selling a local business the recommendation itself. Second, no platform has published a rate card we could read, so we will not print a price. What we will say, as a forecast: we expect prices to rise, as they did on Google and Meta, and we expect the machinery already in ChatGPT’s Ads Manager (geographic targeting, bidding on outcomes) to reach local businesses.
What a zip code is worth
Here is the arithmetic an owner asked us for. It is our scenario, not a platform’s plan, and the assumptions are in the table.
There are 41,554 ZIP codes in the United States.[32] Suppose an AI platform sold one exclusive recommended slot per ZIP code per trade, and suppose the business in that slot paid what a strong local advertiser pays for a channel today.
| Scenario | Who pays (our assumption) | Monthly (our assumption) | The multiplication |
|---|---|---|---|
| Conservative | 1 slot × 41,554 ZIPs[32] × 5 trades = 207,770 businesses | $3,000 | 207,770 × $3,000 × 12 = about $7.5 billion a year |
| Mid | 1 slot × 41,554 ZIPs[32] × 20 categories = 831,080 businesses | $5,000 | 831,080 × $5,000 × 12 = about $49.9 billion a year |
| Aggressive | 2 million to 3 million businesses, a third to a half of all 6.1 million employer firms[36] | $3,000 to $8,000 | $72 billion to $288 billion a year |
The honest verdict. The conservative case is the floor. $7.5 billion a year is about 4% of the $186.1 billion forecast for all US local advertising in 2026,[33] and about 11% of Amazon’s 2025 ad revenue.[16] The mid case is plausible only if AI becomes the default local channel. It would be about 27% of all US local ad spend, or more than half of the digital part, which was $90.4 billion in 2025.[34]
The aggressive case, and anything above it, is not a number we will defend. All US local advertising, every medium and every advertiser, was $169.0 billion in 2025.[34] There are about 6.1 million employer firms in the country, out of 33.2 million small businesses of every size,[36][35] and companies put about 9% of revenue into marketing, 13% to 16% for the smallest.[37] “Millions of businesses paying $3,000 to $8,000 a month” would exceed every dollar spent on local advertising today, and “trillions” would exceed it many times over. The arithmetic is only honest with the assumptions attached, so ours are printed above.
What we may not say
We wrote this page to be quoted, so here is the line we hold.
- We do not guarantee that you will be the recommended business. A tracked business appears in 32% to 38% of AI local answers, and nobody controls the model’s reply.[24]
- We do not know what AI ads will cost a local business. No rate card exists that we could read. “Prices will rise” is our forecast, built on Google’s and Meta’s history.
- Pay-to-play for the recommendation itself is not here. Ads beside answers are. The rest is our forecast.
- Google clicks do not cost ten times what they did. Google’s own average fell from 2012 to 2020 and has risen since 2021.[4]
- Meta’s roughly 20x is a price-per-ad index that includes a format change. Since 2016 it is up about 42%.[10]
- Early adopters do not always win. The academic record on first movers is mixed, and the advantage depends on resources and timing.[38] Our claim is about the price of entry, not a guaranteed result.
What to do
Two things, both free. Run the free AI Visibility Score to see what ChatGPT, Gemini and Google’s AI say about your business today. Then book twenty minutes and we will read the result together and decide whether there is a fit. If the AI is not naming you, the AI visibility page explains what it reads, and the AI receptionist page explains what happens when the call it sends you arrives.