Google AI Mode Is Now Bigger Than Google Gemini
* In March 2025, Google launched Google AI Mode, which is a chat experience powered by Gemini and embedded within the Google search experience.
* Growth plateaued in March 2026 and decreased through April 2026.
* However, in May 2026, Google announced a deep integration with AI Mode and search.
* Since May 2026, AI Mode has increased +76% in only 3 months (through August 2026).
* As of August 2026, AI Mode is now more than 36% larger than Gemini for web and mobile app.
AI Is Growing // State of Generative AI 2026
We published a new study detailing the growth in generative AI. https://t.co/oEVY9hikpN
Key Takeaways
* The top 6 LLMs now drive 76B sessions a month worldwide, and YoY growth has hit 37% worldwide and 45% in the US.
* Headlines have highlighted decreases in referral traffic from AI and AI “stealing traffic.” However, referral traffic from AI continues to climb, with 92% YoY growth worldwide. ChatGPT owns 60% of referral traffic, followed by Gemini at 14%, and Claude at 10%.
ChatGPT
* ChatGPT was the fastest-growing consumer app in history after reaching 100M users in 2 months.
* Starting in July 2026, ChatGPT’s growth has returned, and September 2026 is +24% YoY worldwide.
* Claude has the largest relative growth, with +1901% worldwide and +862% in the US. However, ChatGPT still has an 81.2% share worldwide and 80.7% in the US.
* Despite the growth of Claude, Gemini, and other LLMs, not only has ChatGPT not lost traffic share, it’s gained traffic share +0.8 percentage points from June to September 2026.
Claude
* Claude has seen one of the largest growth trajectories among LLMs.
* US sessions are up +252% between January and March 2026, from 90M to 317M.
Google
* Google’s AI Mode sessions have surpassed Gemini, which shows the convergence of search and AI.
ChatGPT Ads Webinar
Recording: https://t.co/EeCJM72zTK
The Market
* Only 43% of ad-eligible users have seen an ad. So, we're still very early in channel maturity.
* CTR has doubled over the last few months.
* B2B SaaS is one of the largest advertiser categories, with Monday, HubSpot, Zapier, Asana, and more all as top advertisers.
ChatGPT Ads Had A Slow Start
* Early advertisers may have struggled because ChatGPT ads were a new product and the platform was still developing.
* Performance has improved over time, particularly for advertisers testing a wide variety of creatives and landing pages.
* ChatGPT ads require more experimentation than channels where advertisers can rely on automated campaign setups, such as Performance Max and Meta.
Measurement & Attribution
* ChatGPT tracking and attribution have been trending due to issues tracking ad traffic as organic and a drop in GA4. ChatGPT-attributed sessions temporarily fell 45%.
* However, if you tag URLs with the correct params, tracking should work.
The Ads
* Context: Many ads are shown deep into conversations. This means that ads are highly contextual and must also be highly relevant to avoid disrupting UX.
* Copy: Many ad titles are actions (create a resume), use cases (resumes done right), or features (resume creator).
* Landing Pages: Other ad channels often go to the homepage. ChatGPT ads go to a broader set of pages like feature pages, demos, and product pages.
* Unique Ads: The vast majority of ads are not clones of paid search or Meta ad creatives. They are net new.
Experimentation Mode
* The top advertisers have the most unique creative and landing pages. Some have just 1, but others have dozens.
* Jotform: A top advertiser, has almost 1,000 unique ad creatives.
* ChatGPT Ads Playbook: We include our playbook for building ChatGPT ad campaigns.
How Long Does It Take To Impact AEO?
* I have previously said that while SEO takes time (3+ months usually), optimizing for AI can happen faster (days). This is partially true.
* Many AI answers do not use search and do not have citations (10%-80% of answers have citations depending on how you decide which prompts to focus on). Answers without search and citations rely entirely on the underlying model's parametric knowledge.
* According to our analysis of public disclosures from LLM companies, it takes at least 2 to 5 months for AEO initiatives to affect parametric knowledge.
Why Does It Take So Long?
* New models are released every few months. Further, new models have a "training data cutoff" which ranges from a few weeks to a few months old. So, a model released today isn't trained on data released yesterday. There is a time delay to train these new models.
Why This Matters
We recommend complementing strategies based on citations with parametric knowledge optimization (PKO):
* Most AEO strategies focus on citations and ignore the model’s parametric knowledge, the information stored in the model’s weights.
* But not all responses use search, and even when they do, the model’s parametric knowledge can shape both its fan-out queries and how it uses search results to answer.
* Therefore, we should also focus on PKO, which aims to improve how models describe and recommend a brand by influencing the training set.
Measurement
* Rather than only measuring weekly and looking for incremental trends, also look for large discrete jumps among different model versions.
* When you see a change in a new model, check our paper or the LLM to see the training data cutoff and only draw conclusions about changes you made prior to the training data cutoff. Anything after necessarily cannot have an effect until a newer model has it in its training data.
* Use APIs with web search disabled to ensure responses are based only on parametric knowledge. Collect at least fifty responses per prompt. This process is needed only once per pinned model version, since the response distribution will not change as long as the prompts, system instructions, and sampling settings remain fixed.
Parametric Knowledge Optimization & AEO
LLM vs. LLM + Search
* LLM - LLMs are trained on billions of documents to build their models. Parametric knowledge is the knowledge an LLM learns during training.
* LLM + Search - Sometimes, LLMs also perform search or other tasks to add necessary context. For example, if you ask what the weather is, an LLM won't know, but it can retrieve live weather information via search to help answer this question.
AEO Is Not Just Citation Optimization
* Most AEO strategies aim to improve a brand’s visibility, position, and sentiment in AI answers by growing its citation share or earning mentions in cited third-party sources.
* Most AEO strategies ignore the LLM and its parametric knowledge, the information stored in the model’s weights.
Many AI Answers Don't Use Search
* Depending on the data source, search is used between 10%-80%. For many entity queries (best X for Y), search is used often. However, many other prompts don't use search. Similarweb estimates that roughly only 10% of ChatGPT answers have citations.
Winning Citations Isn't Enough
* Even when search is used, being cited is not enough. We present evidence that a brand with low visibility in the LLM without search rarely appears even when it's cited with search (11% of the time). Brands that appear often in the LLM without search appear almost always with search.
* So, much of what drives which brands are mentioned is the parametric knowledge of the underlying model (even when search is used).
Parametric Knowledge Optimization
* For AI answers without search, optimization of search and citations has little to no effect. Instead, you need to optimize the parametric knowledge.
* We need a new set of AEO strategies to optimize for the parametric knowledge.
Measuring Parametric Knowledge Is Hard
* Unlike citations, we cannot directly access the inputs to the underlying model.
* New models are launched every few months, and there is a cutoff to the training data. So, a change made today will take several months to have an effect.
Parametric Knowledge Influences Fan-Out Queries
* Many AI answers use fan-out queries to gather information. Many of these fan-out queries are site:https://t.co/ZDXwHnp0Mt queries, such as -- site:https://t.co/wWnCOA914S best running shoes -- as a fan-out.
* These heavily affect AI answers. The underlying model drives these fan-out queries. To win these answers, you need to optimize the parametric knowledge.
Full Paper: https://t.co/NUmHvY5sEa
How To Gather Real AI Prompts
* There is much focus on how to optimize brands for particular prompts in AI. However, most people guess what prompts they should optimize for. This can call the entire strategy into question.
* Outside of Similarweb, in our opinion, there are no reliable data sources that show users' real prompts. Existing tools are guessing based on dubious underlying data.
* We propose a way to gather real user prompts in AI at scale for free.
* Guide: https://t.co/nxJyVOBmGn
How Did You Hear About Us?
* Attribution in AI is also hard. One solution is to ask users to self-report the channel primarily responsible for their conversion. This is often done by asking “How Did You Hear About Us?”
* Self-reporting can also be extended to gather more information, namely, real user prompts.
What Prompt Did You Use?
We built a new conversion funnel at Graphite that asks the following questions.
1. How Did You Hear About Us? (channel)
2. What LLM Did You Use (product)
3. What prompt did you enter (prompt)
This now tells you not only whether a user entered via AI, but also which LLM they used and the prompt they entered.
Results
* We now see 30 % of inbound leads come from AI. 52% came from ChatGPT and 40% from Claude.
* We receive 10x more referral traffic from ChatGPT than Claude, while the % of leads that self-report Claude is almost as high as ChatGPT. This suggests conversions via Claude are particularly underreported using last-touch, far more than ChatGPT.
* AI leads tend to be less qualified than other channels like referrals.
* Only 21% of leads that self-report as coming from AI appear as referral traffic from AI. The rest came from search (49%) and direct (30%).
Opus 5.5 Is Becoming More Human
* We recently published our AI tells study that looked at ChatGPT, Claude, and Gemini across the previous large model releases and compared their tells against each other and human text.
* Anthropic states that "Opus 5.5 communicates more naturally than prior models." Our Chief AI Officer, Gregory Druck, evaluated these claims and presents data supporting that Opus 5.5 is closer to human writing.
* Study: https://t.co/SJaYftg5Qf
Our Findings
* Claude’s word choice is becoming more similar to human writing, while GPT’s is becoming less similar.
* Opus 5.5’s word-distribution divergence from human writing is 19% lower than Opus 5's. Astra’s is 8% higher than Sol's.
* Opus 5.5 uses less mannered prose (metaphor and flourish in place of direct statements) than Opus 5, but still more than Astra.
* Opus 5.5’s mannered-prose score is 37% lower than Opus 5's. It remains about 1.6 times the human score, compared with 1.2 times for Astra.
* Both Claude and GPT now use well-known tells like the em dash less often.
* Opus 5.5 uses em dashes 99% less often than Opus 5. Astra uses them 95% less often than GPT-5. Both now use them much less often than human writers.
* Opus 5.5 still has many tells. AI tells change between models and versions, but do not go away.
* Opus 5.5’s overall word use is more similar to human writing, but it still overuses many words and phrases. We find 2,548 tells, just 4% fewer than in Opus 5.
* Opus 5.5 uses more language about helpfulness. For example, “can help you” appears eight times as often as in Opus 5. Emphatic phrases like “is genuinely” are less common in Opus 5.5, appearing 26 times as often in Opus 5.
* Opus 5.5 uses more superlatives than Astra. For example, “the most powerful” appears 24 times as often.
The State Of ChatGPT Ads
* We partnered with @Similarweb on a large-scale study of ChatGPT ads
* Data: Similarweb's panel of millions of real people. Graphite analysis of 98k unique ads, 8k landing pages. We are also running ChatGPT ads for several clients and have several case studies to pull from.
* Study: https://t.co/xyOB1S7BOT
The Story
Our overarching thesis is that ChatGPT ads are a huge new channel. They're at $1B revenue run rate in only a few months.
Early advertisers started then stopped. Some people think it's not an effective channel yet. But the data show that many advertisers are growing their ChatGPT ad campaigns, and the ones continuing to advertise are the ones testing and iterating.
The Market
* Roughly half of all users can see ads (Free or Go tier). The other half have premium subscriptions and cannot see ads.
* Only 43% of ad-eligible users have seen an ad. So, we're still very early in channel maturity.
* CTR has doubled over the last few months.
* B2B SaaS is one of the largest advertiser categories, with Monday, HubSpot, Zapier, Asana, and more all as top advertisers.
The Ads
* Context: Many ads are shown deep into conversations. This means that ads are highly contextual and must also be highly relevant to avoid disrupting UX.
* Disruption Tolerance: I suspect that ads must pass a much higher bar because they're shown deep into conversations vs. an ad on Instagram that is between stories. The latter has much less work to be considered relevant by the user.
* Copy: Many ad titles are actions (create a resume), use cases (resumes done right), or features (resume creator).
* Landing Pages: Other ad channels often go to the homepage. ChatGPT ads go to a broader set of pages like feature pages, demos, and product pages.
* Unique Ads: The vast majority of ads are not clones of paid search or Meta ad creatives. They are net new.
Experimentation Mode
* The top advertisers have the most unique creative and landing pages. Some have just 1, but others have dozens.
* Jotform: A top advertiser, has almost 1,000 unique ad creatives.
* ChatGPT Ads Playbook: We include our playbook for building ChatGPT ad campaigns.
We just did this as a one off test for one month to evaluate getting mentions. We don’t do any incentivized / paid for stuff around mentions or anything other than this one time for one month for a few URLs.
I don’t disagree with your argument that incentivized mentions are arguably questionable.
AI-Generated Content Has Plateaued
* We published a new study with @axios evaluating the prevalence of AI-generated content: https://t.co/tQ0GPMoBjG
* AI-generated content grew from the launch of ChatGPT in November 2022 to nearly 50% of all content published on the internet in Q1 2025.
* However, since Q1 2025, AI-generated content has remained flat at 50% and human content 50%.
* Why?: We do not know for sure the cause of the plateau. However, since 2025, there has been increasing evidence that AI-generated content often (but not always) performs poorly in SEO and AEO. We published studies in 2024 and 2025 showing that only 14% of articles in Google Search are generated using AI, and only 7% for position 1 in Google.
* Methodology: We evaluated 55.4k articles randomly selected from Common Crawl between January 2020 and March 2026. We used 3 AI detectors (Pangram, Copyleaks, GPTZero) and show the average of all 3. We describe the methodology in detail in our study.
* AI-Detection Evaluation: We rigorously evaluated each AI detector and showed a low false positive and false negative rate (<2%). We are unable to fully evaluate AI detection for AI + human hybrid content.
* Data Source: Common Crawl maintains one of the largest publicly available web archives. It contains billions of pages and is used by researchers and developers. It is a key data source for training large language models. However, it is not necessarily representative of what real people see and is not an exhaustive snapshot of the entire internet.
* Prior Research: We previously published data as of May 2025 showing 52% rather than 50%. This difference between the two studies is due to using 3 separate AI detectors.
* Oxford: Our previous went viral on social media after Perplexity's X bot published a summary, which hallucinated and credited Oxford as the source. However, Oxford was not involved in either study.
I originally hypothesized that AI was over hyped and the size of its growth curve overstated. So, I sought to evaluate and likely disprove the prevailing narrative about the size of AI. However, what I found is that the size and trajectory of AI are *understated* and that AI is much bigger than I thought.
Study: https://t.co/4LjR9LN1LE
Key Findings
1. Monthly sessions of AI are now 56% the size of search worldwide and 34% in the US, 4x-5x larger than previous reports that only include web data.
2. AI now receives 45B monthly sessions worldwide, and 5.4B monthly sessions in the US.
3. Search-related sessions of AI (Asking prompts) are now 28% the size of search worldwide and 17% in the US.
4. AI has grown even larger across the World than in the US, with worldwide sessions 7x that of the US.
5. Search has not decreased, and neither has Google. Instead, the pie has gotten bigger. Total search-related sessions, combining search engines and search on LLMs, have increased by 26% worldwide and by 16% in the US, comparing 2025 with 2024.
6. 83% of AI usage occurs in mobile apps worldwide, and 75% in the US.
Marketers everywhere are wondering the same thing:
Is our brand appearing in AI responses?
What types of content performs?
What comes next?
Join Graphite CEO @ethan_graphite and @AirOpsHQ CEO @alexhalliday, two leaders shaping the future of marketing, as they reveal how to stay visible and build systems for growth in 2026.
You’ll get clarity on:
- What actually drives visibility in AI search
- How top brands are evolving their content systems
- Real tactics to improve visibility and stay ahead of what’s next
Details:
▶️ Webinar: How to Win the AI Search Shift
🗓️ November 19, 2025 | 4 PM EST
➡️ Reserve your spot: https://t.co/tDnJF3MK5o
@webflow are awesome!
So nice to see them surfing the AI wave in such a brilliant fashion and well accompanied (with @ethan_graphite - where he covers the rise of answer engines, where the future of search is headed, and what you can do today to ensure answer engines cite your website.)
https://t.co/qXWTARd9wS
@tferriss 10. Having an assistant is a huge unlock, but it’s very hard to hire a good assistant + train them. More often than not, assistants create more work for you, not less. Athena is great because this is all turn key. I have one Athena assistant now. Considering hiring a second.
When @ethan_graphite and I first met 8-9 years ago, SEO was simple. Build links, target keywords, fix technical issues and you would rank. Now, search is changing faster than ever — and AEO is the next big shift.
@clearscope, we’ve always believed great content deserves to be found. But as Google and AI evolve, the rules of discovery are being rewritten.
That’s why I’m thrilled to host Ethan Smith, CEO of @graphitegrowth and one of the smartest strategists in AEO & SEO, for our next Clearscope webinar on Monday (10/27) at 2p Central Time.
Ethan has helped companies like Webflow and Robinhood dominate search—and he’s been thinking deeply about how AI and Answer Engine Optimization (AEO) will reshape the way brands earn visibility.
In this session, we’ll dig into:
🔍 What AEO actually is—and why it’s the future of SEO
⚙️ How to adapt your strategy for AI-driven search experiences
🚀 What it means to build content that actually gets discovered
If you care about where SEO is headed, this is one you don’t want to miss.
📅 Save your spot👇