HomeDigital MarketingAI Halftime Report H1 2025

AI Halftime Report H1 2025

It’s halftime.

The primary half of 2025 introduced main shakeups in web optimization, AI, and natural development – and it’s time for a actuality test.

Visitors is down, income is … difficult, and enormous language fashions (LLMs) are not fringe.

Publishers are panicking, and web optimization groups are reevaluating how they measure success.

And it’s not simply the tech shifting; it’s the financial system round it. The DOJ’s antitrust case towards Google might reshape the taking part in subject earlier than This autumn even begins.

In right now’s Memo, I’m unpacking the state of natural development on the midpoint of 2025:

  • How AI Overviews and AI Mode are consuming clicks, and what meaning for TOFU, MOFU, and BOFU content material.
  • Why publishers are suing Google and getting ready for zero site visitors.
  • What’s actually taking place with tech layoffs and job transformation.
  • How we measure LLM visibility right now, and the place that’s headed.
  • What to anticipate subsequent in natural development, search, and monetization.

Plus, premium subscribers will obtain my scorecard that may assist consider whether or not the staff is adapting successfully to the AI panorama.

Let’s take inventory of the place we’re, and what comes subsequent.

Picture Credit score: Kevin Indig

Increase your expertise with Progress Memo’s weekly knowledgeable insights. Subscribe at no cost!

AI Is Reducing Flesh

AI Overviews (AIOs) regarded “fascinating” to entrepreneurs in 2024 and “devastating” in 2025.

The site visitors loss impression ranges from 15% to 45% declines, from my very own observations.

Backside-line metrics throughout the trade vary from “site visitors down, income up” to “site visitors down, income down.”

In February, I wrote in The Impression of AI Overviews that principally the highest of the funnel (TOFU) queries had been impacted:

Each research I checked out confirmed that almost all of AI Overviews present up for informational-intent key phrases like questions.

Shortly after, in March 2025, Google nullified that idea by dialing up the variety of AIOs manner past the highest of the funnel.

Ever since, U.S. firms have skilled a robust (detrimental) impression, and I’m listening to the phrase “web optimization is lifeless” extra usually from leaders.

Between 13 and 19% of key phrases present AI Overviews, in accordance with Semrush and seoClarity, however I assume the precise quantity is way increased as a result of searchers use for much longer prompts. (Prompts that the majority instruments don’t observe.) [1, 2]

I anticipate natural site visitors to maintain dropping because the 12 months strikes ahead.

In theAIO Usability research I revealed in Could, solely a small fraction of clicks nonetheless got here by to web sites.

It wouldn’t shock me if 70% of the natural site visitors that websites earned in 2024 is passed by 2026, leaving simply 30% of that natural site visitors behind.

Scary? Sure. However site visitors is only a means.

The identical research additionally reveals that 80% of searchers nonetheless lean on natural outcomes to finish their search journeys.

So, I nonetheless really feel optimistic in regards to the worth of natural search in the long run.

There are two questions prime of thoughts for me for the time being:

  1. If AIOs actually solely impression the highest of the funnel, then why are income numbers down?
  2. At which level is the decline going to stage off?

For my part, both:

  • AIOs are actually principally TOFU queries. In that case, TOFU content material all the time had extra impression on the underside line than we had been capable of show, and we are able to anticipate the site visitors decline to stage off.
  • Or AIOs impression far more than MOFU and BOFU queries as properly (which is what I believe), and we’re in for an extended decline of site visitors. If true, I anticipate income that’s attributed to natural search to say no at a decrease charge, or by no means for sure firms, since buy intent doesn’t simply go away. Subsequently, income outcomes would relate extra to our capacity to affect buy intent.

With one exception.

Publishers Are Struggling

The entire web is making an attempt to determine whether or not the worth of displaying up in LLMs (ChatGPT, Gemini, AI Mode, AI Overviews, and so on.) is price greater than the loss in site visitors.

However surely, publishers and associates are the group that will get hit the toughest on account of their reliance on advert impressions and hyperlink clicks.

Nobody wants site visitors as a lot as publishers.

Picture Credit score: Kevin Indig

The consequence? Main publishers and information websites will conduct layoffs and assume that Google site visitors will go to zero in some unspecified time in the future.

At a companywide assembly earlier this 12 months, Nicholas Thompson, chief govt of the Atlantic, mentioned the publication ought to assume site visitors from Google would drop towards zero and the corporate wanted to evolve its enterprise mannequin. [3]

Publishers within the EU have banded collectively and filed an antitrust criticism towards Google for its launch and the impression of AI Overviews with the European Fee. [4]

Publishers utilizing Google Search wouldn’t have the choice to decide out from their materials being ingested for Google’s AI giant language mannequin coaching and/or from being crawled for summaries, with out shedding their capacity to seem in Google’s normal search outcomes web page.

I caught up with Chris Dicker, who leads one of many co-signatories within the DMA criticism towards Google, the Unbiased Publishers Alliance:

Kevin: What’s your function within the lawsuit towards Google?

Chris: The Unbiased Publishers Alliance is among the co-signatories on the criticism. I’m serving to lead this from the Alliance aspect.

Kevin: What can be an final result, i.e., an motion by Google, that will be passable?

Chris: We’re solely asking for what we deem to be truthful, which is for a sustainable ecosystem.

Whether or not that’s fee to be used of content material or for Google to begin to considerably scale back the zero-click searches, which have gotten considerably worse for the reason that launch of AIOs.

Kevin: Can LLMs (ChatGPT & Co) present some treatment towards the site visitors drop from Google?

Chris: Not for publishers for the time being, no. They don’t have scale or the need to ship site visitors anyplace else. The present CTR’s we’re seeing and which can be being reported from publishers are tiny.

OpenAI’s scrape to human go to is 179:1, in contrast with Perplexity’s 369:1 and Anthropic’s 8692:1 (stats from Tollbit’s State of bots Q1 2025).

For perspective, Bing’s is 11:1. I do know there are reviews that the site visitors from LLMs is “higher high quality,” however not on the metrics that will assist publishers or content material creators.

It is rather a lot the alternative: Bounce charge is increased; pages per session and per go to are additionally each significantly down for AI search site visitors in comparison with natural search.

Kevin: What are the implications of Google’s AI Overviews on unbiased publishers thus far? Are you able to quantify the impression?

Chris: It’s vital and one thing that has been extrapolated even since April this 12 months. There are websites which can be seeing site visitors drops of as much as 70% since April.

Publishers don’t have any alternative however to chop prices and, sadly, that additionally means job losses.

Within the final 12 months, now we have had quite a few members who, sadly, haven’t been capable of climate the storm and have ceased publishing altogether, and these are revered websites that had been properly established during the last 10 years, if not longer.

Kevin: Have you learnt of publishers which can be capable of dampen the detrimental impression from AI Overviews in some methods? If that’s the case, what are they doing?

Chris: Almost each writer I converse to is actively diversifying away from Google.

It feels inevitable that we’ll see a mass blocking of Googlebot in some unspecified time in the future, one thing that will have been inconceivable simply 12 months in the past.

If your enterprise mannequin nonetheless depends on search site visitors, whether or not from conventional search or AI-powered outcomes, it’s time to rethink – and quick.

Extra publishers are actually specializing in direct viewers relationships by newsletters, boards, podcasts, and related channels.

Platforms like Substack supply an fascinating mannequin, although I’m not satisfied their method absolutely fits publishers simply but.

Past monetizing web sites and content material, many publishers are additionally growing in-house inventive, social, or AI businesses. In any case, these companies have spent years partaking and provoking audiences.

Serving to advertisers faucet into that experience looks like a pure subsequent step.

Moreover the truth that the open internet and demanding societal cases are fading away, from a purely sensible standpoint, there are additionally fewer publishers to amplify content material for different companies.

And but, I imagine we haven’t seen the total extent to which Google Search will change from sending site visitors to answering questions immediately.

AI Mode Is Sitting On The Bench, However It Appears Prepared

At a current occasion I attended, a Google consultant talked about that Sundar Pichai sees AI Mode because the default search expertise within the subsequent two to a few years, with searchers with the ability to swap to traditional search outcomes in the event that they need to – assuming customers like AI Mode.
And that appears to be the case: In keeping with a (small) survey performed by Oppenheimer & Co., 82% of searchers discover AI Mode extra useful than Google Search, whereas 75% discover it extra useful than ChatGPT (I’m wondering why). [5]

Nothing reveals worry greater than copying a challenger’s consumer interface and abandoning the money machine that labored for 20 years.

AI Mode is mainly ChatGPT with a Google emblem. Google follows the Meta playbook, which fenced in Snapchat’s and TikTok’s development by copying their core options.

And most alarmingly for search entrepreneurs, AI Mode eats clicks for breakfast.

Analysis by iPullrank discovered that “4.5% of AI Mode Periods lead to a click on.”[6]

A click on. As in a single!

However Google can’t afford to lose the investor narrative.

I personally imagine that AI Mode received’t launch earlier than Google has discovered the monetization mannequin. And I predict that searchers will see manner fewer adverts however significantly better ones, and displayed at a greater time.

As a result of conversational interface and longer prompts, Google shouldn’t solely have extra context about what customers really need, however they’d additionally be capable of higher estimate when is the perfect time to indicate an advert through the chat dialog.

Consequently, I anticipate CPCs will skyrocket, however CPAs will turn into extra environment friendly.

AEO/GEO/LLMO: Too Many Buzzwords However Not Sufficient Differentiation

Between AI Mode, AI Overviews, and ChatGPT stands this essential query:

How a lot can we affect solutions, and the way totally different is that job from what we’ve performed in web optimization during the last 20 years?

It’s easy. The ways are principally the identical, however the ecosystem adjustments:

1. Longer prompts: The typical immediate is 23 phrases lengthy in comparison with 4.2 for traditional Google Search. [7]

The wealthy element customers present about their intent hits a content material hole that’s tuned for shorthead key phrases on the opposite aspect of {the marketplace}.

Consequently, I see hyper-specialized content material that’s fine-tuned for particular personas (see How one can Optimize for Matters) in our current and future.

2. web optimization winners usually are not AI winners: If web optimization was sufficient and there was nothing else we would have liked to do “for AI,” then why aren’t the websites which can be most seen in Search the identical ones which can be seen in LLMs?

In Is GEO/AEO the identical as web optimization?, I discovered that the lists differ drastically in most verticals. Solely extremely consolidated areas with a number of winners, like CRM software program, have an identical winners throughout each modalities.

3. New intent: Generative: Semrush and Profound got here to the conclusion that ~30-70% of intent on LLMs is “generative,” which means customers need to accomplish duties proper then and there. [8]

What’s usually missed is that whereas performing an motion, e.g., producing a picture, the intent can shortly flip to informational or transactional, e.g., study extra in regards to the subject you need to generate the picture about or purchase icon license.

Since experiences are conversational and extra steady, we have to replace our mannequin of intent. It doesn’t occur in isolation (suppose: one session), however a number of intents can happen throughout the identical session (informational → generative → transactional → informational → and so on.).

My opinion: It’s too quickly to coin a time period.

Will we swap from Reply Engine Optimization to Agentic Engine Optimization after we enter the Agentic AI age? AI has advanced at a rocket tempo during the last 2.5 years, and I don’t anticipate it can decelerate quickly.

LLMs Are No Longer Fringe

In 2025, LLMs reached the mainstream. We’re not speaking a couple of fringe platform anymore: ChatGPT supposedly receives 2.5 billion prompts a day.
With Google seeing over 5 trillion searches per 12 months, you can say ChatGPT has reached about 17.8% of Google’s quantity.

Needless to say loads of prompts usually are not searches on ChatGPT, after which the comparability turns into weaker (till Google rolls AI Mode out broadly). [9]

Picture Credit score: Kevin Indig

Essential to notice is that LLMs depend on totally different quotation sources to various levels. [10]

Profound noticed in 30 million citations that ChatGPT, AIOs, and Perplexity depend on totally different quotation sources:

  • ChatGPT cites Wikipedia virtually 50% of the time, adopted by citing Reddit at 11.3% and Forbes at 6.8%.
  • AI Overviews cite Reddit 21% of the time, adopted by 18.8% for YouTube, 14.3% for Quora, and 13% for LinkedIn.
  • Perplexity cites Reddit virtually 50% of the time, YouTube at 13.9% of the time, and Gartner at 7%.

We all know that investing time and assets into non-Google platforms is vital to constructing belief and visibility throughout all platforms.

However now we all know that the combination of platform funding is dependent upon the place you need to construct visibility.

Reddit appears to offer common impression, which is smart given their licensing offers with OpenAI and Google, however YouTube, Quora, and assessment platforms don’t present the identical potential for gaining citations on all LLMs.

Picture Credit score: Kevin Indig

Time additionally issues. AirOps discovered that 95% of pages cited in ChatGPT are lower than 10 months outdated. [11]

An enormous cause for that is the coaching knowledge cutoff for LLMs. New fashions are nonetheless educated on giant corpi of information (bear in mind the Google Dance?).

Something newer than the time of coaching wants to return from the net. Consequently, holding content material recent and repeatedly iterating looks like a path to AI visibility to me. Even including the present 12 months to the URL (and meta-title) looks like a good suggestion. [12]

A research by Apple, which I coated within the Progress Intelligence Temporary, raises a query we’d all have on the tip of our tongue: Are LLMs overhyped? [13]

The reply: It relies upon … on the complexity of the duty:

  • Easy issues: Fashions usually discover appropriate options early however wastefully proceed exploring incorrect ones (“overthinking”).
  • Reasonable complexity: Fashions discover many incorrect options earlier than discovering appropriate ones.
  • Excessive complexity: Fashions fail to generate any appropriate options.

LLMs are good however nonetheless battle with complicated duties. Excellent news for tech employees … proper?

And right here’s one other factor: With the rise of LLM use and adoption, how will we measure success for our optimization efforts?

I ran a survey of Progress Memo in June, and it’s clear our trade hasn’t actually nailed how we measure the LLM visibility of our manufacturers.

Out of those that responded, about 30% are utilizing conventional web optimization instruments to measure LLM visibility, 26% are utilizing Google Analytics 4 site visitors alerts, and a whopping 21% aren’t measuring but and need assistance figuring out how.

Picture Credit score: Kevin Indig

And the largest shock is that this: Overwhelmingly, we don’t belief our LLM visibility measurements.

Near 80% of survey respondents don’t imagine the best way they’re measuring LLM visibility is correct.

Picture Credit score: Kevin Indig

An enormous subject in the entire LLM dialog is, after all, whether or not AI replaces white collar employees or not.

I’m together with this dialogue in my halftime report as a result of I’m seeing a rising variety of in-house consultants who’re afraid to get replaced.

Amazon’s CEO, Andy Jassy, wrote a public memo, saying the corporate would want fewer folks due to AI (bolded textual content is mine):

“As we roll out extra Generative AI and brokers, it ought to change the best way our work is finished. We are going to want fewer folks doing a number of the jobs which can be being performed right now, and extra folks doing different sorts of jobs. It’s arduous to know precisely the place this nets out over time, however within the subsequent few years, we anticipate that it will scale back our whole company workforce as we get effectivity beneficial properties from utilizing AI extensively throughout the corporate.” [14]

Amazon has lower +27,000 jobs between 2022 and 2023, however has by no means had extra staff on the finish of 2024, apart from on the finish of 2021 by a small margin. [15]

Different tech firms pulled even:

  • Salesforce’s CEO, Marc Benioff, says that 30-50% of the work at Salesforce is finished by AI. [16] Salesforce eradicated ~1,000 roles this 12 months.
  • Klarna’s CEO first introduced that AI is doing the work of 700 customer support brokers and fired about 2,000 staff, however then backtracked and rehired people. [17]
  • Microsoft lower 15,000 jobs in 2025. CEO Satya Nadella mentioned AI writes ~30% of latest code in some tasks.
  • Meta laid off 3,600 staff in 2025, with Mark Zuckerberg saying AI might be able to be a mid-level engineer this 12 months.

However is AI actually changing white collar employees, or is it used for good PR?

Layoff tracker, layoffs.fyi, reveals that the variety of firms and staff laid off shouldn’t be rising for the reason that pandemic.

Picture Credit score: Kevin Indig

A jobs report by CompTIA reveals that whereas tech employment is barely down between June 2023 and June 2025…[18]

Picture Credit score: Kevin Indig

…the variety of job openings with AI expertise far outpaces the variety of listings for all roles.

Picture Credit score: Kevin Indig

In different phrases, “AI layoffs” appear extra PR play or justification for job cuts.

However upskilling with AI is vital.

Google Lawsuit Rushes Towards A Closing Resolution On Labor Day

The landmark lawsuit towards Google for being an internet search monopoly concludes by Labor Day (September 1). The DoJ asks for:

  • A compulsory divestiture of Chrome inside a specified timeframe.
  • A five-year prohibition on Google proudly owning any browser.
  • Termination of unique default agreements.
  • In depth knowledge sharing necessities.
  • The correct to hunt Android divestiture if behavioral treatments show inadequate.

Google, however, agrees to finish unique agreements, so we all know Google and Apple will divorce, however opposes a Chrome divestiture and knowledge sharing mandates.

The treatment ruling might have vital implications on the AI race, and the place entrepreneurs ought to place their cash.

For instance, a Chrome divestiture might considerably set Google again, as OpenAI and Perplexity launch their very own browsers. It could additionally imply a fabric loss in consumer habits knowledge and agentic AI capabilities.

Dropping the unique settlement with Apple might additionally imply that extra customers set different browsers than Chrome as default, if they’ll present a robust profit.

Nevertheless, I personally suppose essentially the most sensible final result is a compelled finish to unique agreements and can be shocked to see a Chrome divestiture.

For context:

  • The Division of Justice has achieved two landmark antitrust victories towards Google in 2024-2025, with federal judges ruling the tech large operates unlawful monopolies in each on-line search and digital promoting expertise.
  • Each circumstances have now superior to treatment phases the place courts will decide whether or not to interrupt up components of Google’s enterprise, representing essentially the most aggressive authorities intervention in Massive Tech for the reason that Microsoft case 25 years in the past.

Outlook For H2

The second half of 2025 will doubtless be outlined by adaptation fairly than resistance.

Firms that succeed can be people who foster belief past Google, construct direct viewers relationships, and upskill groups in AI.

Right here’s what I anticipate for the second half of the 12 months:

Accelerating Visitors Decline

  • Natural site visitors losses will doubtless intensify as Google expands AI Overviews.
  • Publishers ought to put together for additional 20-30% site visitors declines.
  • The “new regular” of 30% of historic site visitors by 2026 might arrive earlier than anticipated.

AI Mode Launch

  • Google will doubtless roll out AI Mode extra broadly, however cautiously.
  • Count on a heavy give attention to monetization testing earlier than broad launch.
  • Watch for brand spanking new advert codecs optimized for conversational search.

Writer Adaptation

  • Extra publishers will actively block Googlebot.
  • Elevated give attention to direct income streams (newsletters, memberships).
  • Potential consolidation as smaller publishers battle to outlive.

Measurement Evolution

  • New instruments particularly for measuring LLM visibility will emerge.
  • Trade will begin standardizing on key metrics for AI efficiency.
  • Larger emphasis on income vs. site visitors as success metrics.

Market Restructuring

  • DoJ ruling might reshape the search panorama.
  • Count on new search entrants to realize traction.
  • Browser wars might reignite with AI-native choices.

Featured Picture: Paulo Bobita/Search Engine Journal

RELATED ARTICLES

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Most Popular