You’re not angry at AI content.
You’re angry because the slop is winning. And the slop isn’t even trying.
On March 13, 2025, Google’s March Core Update dropped and in the eight weeks that followed, AI Overviews more than doubled. Ahrefs analyzed 25 million of them and found a 116% jump in volume, plus the share of US keywords that trigger an AI Overview doubled from 7.6% to 16.48%. By March 2026, AI Overviews appeared on roughly 48% of all Google search queries globally up from just 6.49% a year earlier. That’s the surface area where your “high-quality” content is being measured.
But here’s the part that should make you put the phone down.
When an AI Overview is present, the page sitting in position #1 your position #1 gets a 34.5% lower click-through rate than it would have without the Overview. That’s not theory. Ahrefs data scientist Xibeijia Guan pulled the numbers from 300,000 keywords and a like-for-like comparison against March 2024 (pre-AI-Overviews) vs. March 2025. CTR for the top result dropped from a forecasted 4.0% to an actual 2.6%.
So while you were arguing in Slack about whether AI content “deserves” to rank, Google shipped a feature that punishes the page any page that used to own the click.
And the clicks didn’t go to better content. They went to a 6-paragraph machine summary that sourced Reddit, Quora, and your competitor’s worse cousin. Ahrefs’ post-March-Core-Update data shows Reddit’s AI Overview market share jumping 4.2 points in eight weeks more than any domain on earth while established editorial sites like Healthline (-2.6%), Cleveland Clinic (-3.6%), and Wikipedia (-4.7%) lost ground.
You don’t hate AI content. You hate that the system is now optimized to replace you, and the only winning move is one most publishers refuse to make.
The 2026 number nobody is putting on a slide
Let me say the thing out loud that every “AI content” Twitter thread avoids.
As of March 2026, AI Overviews appear on roughly 48% of all Google search queries. They appear on 16.48% of US keywords, and they eat 11.8% of all US search volume. On AI-Overview queries, the top organic result’s CTR is 34.5% lower than it was before the rollout. Independent confirmation from Semrush’s December 2025 study of 10M+ keywords corroborates the trajectory: AI Overviews triggered for 6.49% of queries in January 2025, peaked at 24.61% in July, and settled at 15.69% in November.
Two reputable SEO datasets. Two separate methodologies. Same direction.
Now layer a Pew Research finding reported by The Decoder: only 1% of users click a source link directly from the AI Overview itself. The other 99% get the answer, leave, and never even know your URL existed.
That’s not a ranking problem. That’s a visibility extinction event.
The part that actually hurts your traffic
Look you’re not wrong to be furious. The data backs the rage.
When Penske Media Corporation the parent company of Rolling Stone and Billboard sued Google over AI Overviews in September 2025, they alleged that “20% of searches that link to Penske-owned websites show AI Overviews” and that the figure was rising. Reuters and The Verge both covered it. Earlier, Chegg sued Alphabet in February 2025 over the same issue. The publisher class is no longer pretending they’re litigating.
But lawsuits don’t restore traffic.
What does?
The reason you hate AI content is the reason you should be writing like it
Here’s the unsayable bit.
HubSpot’s 2026 State of Marketing report found that 80% of marketers now use AI for content creation, and 61% believe marketing is experiencing its biggest disruption in 20 years. Ahrefs’ own 879-marketer survey found that 87% of content marketers use AI to create content and that AI-using teams publish a median of 17 articles a month vs. 12 for non-AI teams a 42% output gap.
The competition isn’t “human vs. AI.” The competition is “AI-assisted team shipping 17 pieces a month” vs. “you, handwriting 4, waiting for inspiration.”
You don’t hate AI content because it’s soulless. You hate it because it plays a tempo you haven’t accepted yet. And while you agonize over voice and tone, the AI-only factory is shipping the 1,001st piece of “Ultimate Guide to X” content that ranks for a 4-word query nobody types with feeling.
In April 2026 alone, Ahrefs found that 99.2% of AI Overview keywords are informational in intent, with an average Keyword Difficulty of just 12. You need roughly 13 referring domains to enter the AIO SERP, versus 41 for non-AIO queries. The moat you thought existed? It was always shallower than the conference circuit promised.
The “Google hates AI content” myth that needs to die
I want to kill this with one paragraph.
In March 2024, Google announced new spam policies targeting “scaled content abuse.” Originality.ai’s follow-up study found that 1,446 sites received manual actions on March 5, 2024, and of the 14 publicly disclosed sites analyzed, 100% showed signs of published AI content; 7 of the 14 had AI content on over 90% of their pages. WIRED reported on the same Google crackdown at the time, citing Google’s claim that the update would reduce “low-quality, unoriginal content” in results by 40%.
Read that again. Scaled, low-quality, unoriginal AI content was the target. Not AI content generally.
But here’s what nobody writing headlines about that update told you: Google’s own official guidance says there are “no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary”. The same Google that deindexed 1,446 spam sites in 2024 also published a 2025 study update titled “AI content is not bad for SEO and it never will be” (a position Ahrefs’ own Director of Content has argued consistently).
The rule is real and ruthless:
“Is this spam?” = removed. “Is this helpful?” = ranked, cited, sometimes surfaced in the Overview.
Your position #1 doesn’t care who typed the words. It cares whether the words answered the question faster and more completely than the other 14 results. Per Google’s own Search Central documentation, AI Overviews use a “query fan-out” technique issuing multiple related searches across subtopics and data sources to identify “a wider and more diverse set of helpful links” than classic web search. That diversity pull is your only consistent in.
If the AI Overview cites 8 sources and one of them is your post, you got 12.5% of the credit and almost none of the click. Pew’s analysis says 1% click-through from the Overview. The NYT summarized the dynamic back when Overviews launched: “Google’s A.I. Search Leaves Publishers Scrambling.”
You’re not being demoted. You’re being quoted.
So what actually works in 2026
Everything I’ve read, watched, and tested across the past eighteen months keeps pointing at the same handful of moves.
1. You have to publish more, not less. Counter-intuitive, I know. Ahrefs found 76% of AI Overview citations pull from the top 10 organic results so the old “one perfect post per quarter” play is dead. You need enough surface area that something you wrote gets cited by the Overview at least 40% of the time. That requires volume. AI-assisted teams publishing 17 posts/month are playing the right volume game; the mistake is treating AI like a writer instead of a research assistant and an editor.
2. Each post must contain what Ahrefs’ Director of Content calls “information gain.” The patent analysis Rich Sanger did on AI Overviews found that Google seeks out diversity “If the top-ranked content for that query is homogenous, it will move on to closely related queries.” Translation: restating the consensus answer earns you nothing. You need either experimentation (real data nobody else has), experience (you actually did the thing), or effort (something better than words on a page). Pick one per post. Bonus if you pick all three.
3. The post must do one thing AI summaries cannot: leave the reader with an action they cannot perform inside the Overview. This is the under-appreciated lever. AI Overviews resolve informational intent. They cannot book the restaurant, sign the BAA, install the plugin, or finish the meditation. Posts that move the reader from answer to next step checklist, decision tree, screenshot walkthrough, “if-then” routing earn the click that the Overview can’t steal.
4. Brand and topical authority are now compounding SEO assets, not vanity metrics. Semrush’s 2025 study shows Reddit and Quora dominating AI Overview citations in many categories not because they’re authoritative, but because they accumulated topic density faster than editorial publishers did. Ahrefs independently confirms the same trend Reddit’s AI Overview market share jumped from 1.3% to 5.5% in eight weeks. Every brand needs a Reddit strategy, an original-research pipeline, and a YouTube demo cadence. The sites that show up in three formats on the same topic win.
5. Disclose where you use AI. Strategically. Only 16% of AI-using content teams disclose AI involvement, per Ahrefs but the disclosure isn’t a Google ranking signal. It’s a trust signal. Readers bounce on AI-flavored phrasing faster than any algorithm can catch it. Play the long game. Use AI for research, outlines, first drafts, and citation hunting. Use your scars for the final voice.
What nobody wants to admit
You don’t actually want AI content to stop ranking.
You want a world where your inconvenient craft of writing is rewarded again where a Tuesday afternoon spent sweating a single paragraph beats a Tuesday afternoon spent shipping 4 mediocre posts. That’s a beautiful preference. It’s also not the world we live in.
The world we live in is one where AI Overviews triggered on 24.61% of all queries in July 2025 alone (Semrush), where a German court declared in 2026 that AI Overviews are “Google’s own words” and made the company liable for false answers, where Google restricted AI Overviews on certain health searches after The Guardian found “dangerous” errors in January 2026, and where the EU opened a competition-law investigation in December 2025.
In that world, the question isn’t “does AI content deserve to rank.” The question is “what do I write that Google still has to send a human to read?”
Here’s my answer, and it’s not comfortable.
Write the thing AI cannot finish. The thing that requires a scar, a name, a number that isn’t in the Wikipedia citation. Write the “I lost $14,730 to a contractor and here’s the exact clause that would’ve saved me” post. Write the “I ran this 4,200-word A/B test across 11 sites for 7 weeks” post. Write the “I interviewed 27 CMOs about why they fired their last SEO agency” post.
Those posts rank. They get cited. They get paid.
Not because the algorithm learned to love craft.
Because craft is the only thing left that isn’t infinite.
