Tanking SEO traffic can feel hopeless. I have no doubt AI reduces SEO traffic, but some data I just analyzed makes me believe the story is much more nuanced.

Exclusive data from Similarweb containing the traffic channel mix of 10,000 domains in both the US and France over the last 3 years shows us a 30,000 foot view of how the traffic channel mix changes over time. Other studies have looked at similar angles, but I haven’t seen one that compares countries and spans 3 years.

Data & methodology

Dataset:

• The data is based on Similarweb’s monthly traffic estimates for 10,000 domains in each of the US and France, covering September 2023 through August 2026. The countries contain different website lists, and coverage varies by month, so these samples do not represent the entire national web.

• The dataset contains 663,861 country-domain-month records and 17,863 unique domains across ten channels: organic search, paid search, Direct, referrals, organic social, paid social, email, display, AI and affiliates.

Methodology:

• I compared traffic volumes and channel shares while holding website lists constant within each comparison, then checked sensitivity to site selection, missing estimates and the largest domains.

• Most channel comparisons use January–August 2026 versus the same months in 2025; the two-year organic-search headline compares August 2024 with August 2026 across 8,658 US sites and 7,730 French sites with complete organic estimates throughout the dataset.

• Industry breakdowns use provisional analyst classifications.

• The findings describe modeled visits, not unique people, spending or revenue, and do not establish whether AI or paid search caused organic-search losses.

Organic is down, but less than I expected

Visits from organic search are down -13.2% in the US (19.9b -> 17.3b) and -24.4% in France (4b -> 3b) over the last 24 months (August 2024 -> August 2026).

Another way to look at this is to ask how the share of organic from total traffic has changed during that time. The answer: US organic traffic dropped from 20.5% to 18.3% since September 2023 (-10.8%). In France, it dropped from 27.3% down to 23.1% (-15.2%).

Not a great prospect for SEO, but the decline is admittedly slower in aggregate than what I see in Search Console or GA4 on the site level. The reason is that the aggregate view here in this Memo simply combines websites with very different outcomes. Only when we segment the aggregate can we see the patterns.

3 forms of concentration

Any aggregate metric across a vast sample set usually holds concentration. You know, Pareto ‘n stuff. Losses are widespread across sites, but concentrated in traffic volume. 62.5% of US sites and 66.6% of French sites declined; about 1% of domains account for 40.2% / 48.5% of gross lost visits.

The first form of concentration I want to show you is large vs. small sites. Smaller sites have lost substantially less organic traffic than large ones. The gains are still small in the US (+2.7%) and small sites shrank in France (-1.9%), but there is a big gap compared to large sites. I didn’t use a particular threshold here but simply cut the set of domains into quartiles.

When we use the sumtotal of visits instead of median, we see an even starker picture: small sites gained organic traffic in the US (+19.3%) and France (+42.2%), while large ones lost (-13.5% and -27.4% respectively).

The typical small site was roughly flat: up 2.7% in the US and down 1.9% in France. Their combined traffic grew much faster because a minority of sites made unusually large gains. Remove the 100 biggest small-site gainers in each country, and that combined growth turns negative.

Another form of concentration happens around business models. Here, we see publishers getting hit the hardest because their product, content, directly competes with AI answers. In this sample, publishers lost 41.5% of organic traffic in the US and 42.5% in France over the last 2 years. Certainly devastating losses.

When we strip publishers out, we see an even flatter decline: US down 7.8%, France 22.5%. That’s almost half of the US decline.

The 3rd form of concentration is around topics. Health, food, news and adult-focused sites decline faster than the average in the US and mostly in France as well, with the exception of news. Entertainment and gaming declines a lot slower in the US.

Organic search shrinks unevenly across domain sizes, business models and industries. The median site in the smallest quartile is roughly flat, while publisher traffic has fallen about 42% in both markets. Strong gains at a minority of small sites lift their combined traffic totals. The country average provides context, but comparable websites offer a more useful benchmark for an individual business.

Country-level differences disappear when accounting for mix shift

I compare the US against France because I wanted a control mechanism for AI Overviews and regulation. Google just launched AI Overviews in France in July 2026 and European legislators have forced Google to avoid some SERP Features that impact clicks.

Why is France losing organic traffic faster, especially given Google only rolled AI Overviews out there in July of this year? It’s mostly a matter of site selection. When we compare ony the same sites in both countries the gap narrows substantially.

Across the same 1,725 sites, average organic visits per day fell 5.1% in France and 3.6% in the US between June and August 2026. France weakened more around the launch, but this comparison cannot isolate how much of that change came from AI Overviews.

The same happened when AI Overviews launched in the US: while organic traffic in France was up 2.4%, it was down -0.6% in the US (keep in mind Google rolled AIOs out slowly in the US).

It’s not all AI

How can it be that Google still makes so much money from ads ($63.3b,+16.8% YoY, in Q2 2026) while AI answers a lot of questions? 2 reasons!

One, because Google shows ads before it shows AI Overviews. Two, search ads take up a lot more space and are harder to distinguish from organic results (my opinion - wonder if you agree?).

• On Oct 13, 2025, Google started grouping up to four text ads under a sticky “Sponsored results” header with a “Hide sponsored results” button. It rolled out worldwide on desktop and mobile, and Google said at the time that ad size would stay the same. https://blog.google/products/ads-commerce/google-search-sponsored-results-label/

• Tom Capper from Stat found that since June, the average organic result sits below the common desktop viewport, “aka under the fold”, and ad blocks have doubled in size in a sample of 46k head terms. https://getstat.com/blog/paid-search-getting-bigger-how-to-respond

• Google reports paid clicks up 13% and average CPC up 3% in its Q2 2026 earnings.

In other words, ads eat into organic clicks as well and I think we massively underestimate the impact and misattribute it to AI Overviews. The Similarweb data shows clearly that paid search traffic grows while organic tanks:

The chart above indexes the baseline in August 2024 at 100 and shows the change over time. In the US, paid search traffic has gained 21% and organic lost 13.2%. In France, paid gained 17.3% and organic lost 24.4%.

But, paid search growth doesn’t completely offset organic losses. If I were to take just the absolute traffic numbers from the chart below, paid makes up about 36.6% of organic losses in the US and 6.7% in France. The reality is probably a lot more nuanced and differs by industry, business model and time comparison.

I also find it important to call out that Google’s algorithm updates are quite aggressive. For example, the unnamed January 2026 update wiped out tons of ranks for some sites, and I think this is somewhat visible in the aggregate organic traffic chart from the beginning. Some algo updates lead to a different site type mix, meaning Google might prefer to show more personal blogs rather than large aggregator sites, as I’ve recently seen it for many travel queries.

AI traffic is still a drop in the bucket

Spoiler alert: AI referral traffic has the strongest growth rate of all channels and reaches 2.85% of SEO traffic in the US and 2.44% in France. In other words, AI referrals account for roughly one in 200 visits.

And yet, it’s a drop in the bucket, even if AI traffic converts much higher than organic. AI traffic reaches only 0.5% of total visits. Even if it doubles or triples over the next 12 months, the impact is negligible.

We’re also seeing forms of concentration in AI traffic. Some industries get a lot more than the average. For example, in both countries:

• technology, civic services and careers outperform

• Entertainment, food, sports and adult underperform

• Real estate does better in France, Health does better in the US

I have low hopes that AI referral traffic could ever offset loss of organic traffic. Last week, I wrote about how hard it is to detect hidden AI traffic. But, to be clear, my opinion is that AI traffic is a red herring. I’d focus on whether your brand appears when buyers research your category, then connect that visibility to business outcomes. And just as we’ve seen with traffic, the benchmark needs to fit your industry and market.

This week’s sponsor is Semrush, whose latest AI Visibility Index helps you make that comparison. The study examines which brands and sources appear across four AI platforms in the US and UK, using seven months of data across 22 industries. It shows how those patterns differ between markets, giving you a starting point for deciding where to focus your visibility work.

Get the free study

Don’t pour the baby out!

The way I make sense of all this data is this:

1. Some decline of organic traffic comes from people using AI Chatbots like ChatGPT or Gemini to solve tasks they previously used Search for. For example, our AI Mode study shows an average zero-click share of 64%. So, some organic traffic loss could be simply people searching less (this analysis didn’t look into causality behind this claim).

2. AI Overviews reduce clicks but not as dramatically as AI Chatbots. Users still click on blue links to some degree even when Google shows AI Overviews in our AI Overviews study.

3. Higher ad load on Google and more aggressive algorithm updates cause severe damage to SEO clicks that are often underreported because it’s easier to explain drops with AI.

In some cases, like publishers, the SEO traffic loss is clear, global and painful. In other industries and even countries, it’s a lot more nuanced. Here is the big problem: A lot of people pour the baby out with the bathwater by adopting the simplistic mindset of “AI is killing all of our traffic and SEO is dead”. That might be true in some cases, but often people miss impact from advertising and algorithm updates. In other words, most companies can remedy some SEO click loss. Whether the unit economics still make sense is a different question.

The CEO of a large company recently used an analogy I like: SEO is like a melting ice cube. How fast it melts is not completely out of our hands.

Humbly,

Kevin

Read more

Wouldn’t it be such a relief if we found out there’s actually a meaningful amount of AI traffic that’s just hidden in our analytics tools? I sense a strong belief in the industry that a lot of AI traffic hides as, or arrives as, Direct in our analytics tools. I get where this comes from because I wrote about it:

• In our AI Mode user behavior study, 18% of exits involved users leaving AI Mode and going directly to another site.

• In Is AI cutting into your SEO conversions?, I described a potential journey from AI answers to brand recall to a direct visit, bypassing clicks.

• In How to plan 2027 when traffic is down and AI won’t attribute, I cited Similarweb research showing that users who received AI brand recommendations were more likely to visit those brands’ websites over the following seven days. Most of those visits came through Search (55.9% vs. 40.4% for other visits), and fewer through Direct (19.9% vs. 38.8%).

There’s a plausible mechanism here, but is the signal strong enough to trace Direct traffic from AI in our analytics? Somewhat. And now, we have an example.

In an interview on the Siege Media Podcast, Growth expert Eric Wu discussed how Ramp identified 30–40% of its direct traffic as originating from answer engines. I caught up with him and learned his team examined growth in Direct visits to long blog URLs that users were unlikely to type, filtered out patterns resembling email campaigns, and compared the remaining visits with organic search traffic. Some of that Direct traffic closely matched search visitors in time on page, scroll depth, and click behavior. Self-reported attribution confirmed that the users indeed came from AI.

That got me curious about whether the time is ripe to detect signal in our analytics. So, I went hunting for proof of hidden AI traffic.

Bots. Why did it have to be bots?

I analyzed the Google Analytics 4 properties of 12 sites, using the same August–September 2026 comparison and excluding homepages, localized homepage roots, and unknown landings. Keep in mind that this small sample of 12 sites skews toward B2B, even though traffic is almost evenly split between B2B and B2C. All sites except 2 target the US.

At first, I got really excited: 8 out of 12 properties showed direct traffic growth from August to September! Some even over 100%. But when I dug deeper, much of that direct traffic turned out to be automated bot traffic. At one website, GA4 recorded 409,000 Direct sessions, of which only 198 counted as engaged.

Comparing the increase in all Direct sessions with engaged sessions alone shows a clear picture: engaged Direct traffic has been flat since the beginning of the year. Across the filtered January–September sample, total Direct rose 140.9%, while engaged Direct changed −1.0% in monthly counts, or +2.3% in average daily volume (when normalizing for days per month).

Conclusion: I could not yet detect “hidden” AI traffic without self-reported attribution like Eric and his team at Ramp. I also found that, on average across the 12 sites, Direct sessions with identifiable AI traces are only 0.003% of total direct traffic.

To uncover bot traffic, break Direct traffic down by day, landing page, country, browser, operating system, browser version, and screen size, and look for sudden growth concentrated in groups of visits with the same browser configuration and low or no engagement.

To find AI traffic, I excluded pages that typically get a high degree of direct visits, like pricing or login pages:

• Main homepages, regional root domains, index aliases, and untracked landing URLs.

• Subscription plans and pricing pages, including localized variations and donor-fee breakdown pages.

• Authentication portals, user registration, account management, dashboards, webmail, and internal app paths.

• Product demo requests, help docs, corporate about/team pages, investor relations, privacy policies, and internal search results.

• Transactional flows, payment processing, purchase confirmations, billing receipts, and automated system alerts.

I’ve got a bad feeling about this

I didn’t want to give up here. So, I ran a separate analysis of four brands to test whether greater AI visibility was followed by more branded search or Direct traffic. Across 632 tests, I compared weekly AI visibility with both measures at delays ranging from the same week to six weeks later.

The first results looked promising! One site showed a correlation of 0.72 between AI Share of Voice and Direct traffic, peaking at a three-week delay. Its correlation with branded search impressions reached 0.59. A second site showed a 0.47 correlation between ChatGPT mention rate and Direct traffic. Let’s go!

For the first site, the correlation peaked three weeks later, which made a delayed response look plausible.

Buuuuut a strong correlation needs a meaningful comparison. Two lines can move together because they share a trend, even without one driving the other. Correlation isn’t Cau… You know.

To check, I shifted the AI series against the traffic series, deliberately pairing different weeks while preserving the sequence within each series. If the actual timing carried useful information, it should outperform these mismatched versions. My initial comparison gave the real data an advantage. I had tested seven delays and kept the strongest correlation, but tested each mismatched version only at that winning delay. So I repeated the test. With every mismatched version getting the same seven chances and keeping its strongest result.

For the two Direct traffic examples:

The correct dates beat most alternatives. But several wrong alignments produced equally strong or stronger correlations once they got the same opportunity to search. The relationships also depended on the period. When I split each timeline in half, the first site’s relationship was concentrated in the second half. The second site’s was concentrated in the first. Neither showed a consistent relationship across both periods, although the shorter windows made those estimates less precise.

At the selected delays, these examples had only 17 and 35 usable weekly observations. That limited my ability to distinguish a modest effect from ordinary weekly variation.

The 0.72 correlation looked like the golden idol, but it set off a boulder. Across 632 tests, I compared different brands, AI metrics, and traffic measures. That creates plenty of opportunities to find something impressive. Across the four brands, I couldn’t establish a reliable link between AI visibility and subsequent branded search or Direct traffic. An effect could still exist. I just couldn’t isolate it.

Now, we can still benchmark which brands AI platforms mention and cite, and how that visibility differs by market.

This week’s sponsor, Semrush, compares seven months of US and UK data across 22 industries and four AI platforms in its latest AI Visibility Index study. It shows which brands lead in each market and why a strategy built for one country won’t necessarily carry over to another.

Use the study’s practical benchmarking plan to compare your brand’s visibility in the markets you serve.

Get the free study →

Indiana, let it go

Clicks as underlying basis for attribution exist mainly as a result of availability bias. When marketing measurement evolved from exposure (scores and ratings) to brand lift (“half my advertising budget is wasted”) and eventually clicks, modern tools reporting clicks and using last-click attribution shaped our way of thinking. It also helped that initial CTRs and conversion rates were extremely high. But now that AI breaks the traditional click funnel (search -> click -> convert), we need to rethink.

I found that the amount of Direct traffic to longtail pages in my sample driven by AI is not detectable yet. We also can’t yet find a reliable relationship between AI Visibility and branded search or direct traffic. Some companies with outstanding AI growth might be able to detect this, but across 12 sites of different sectors and sizes, I couldn’t see it.

The better way to measure the impact of AI Visibility is a mix of self-reported attribution and triangulation of leading indicators. In The usefulness of attribution for AI Search, I argued for combining signals with different blind spots:

1. Where your brand appears. Track mentions and recommendations across a consistent set of relevant prompts, platforms, and markets.

2. How people discovered you. Ask new leads and customers where they first heard about you and what prompted them to visit. Capture their own words, including which AI tool they used and what they were researching.

3. What happens commercially. Connect those answers to qualified leads, purchases, and revenue. Follow the people who report AI discovery through the buying process.

As Amanda Natividad and I discussed in How to plan 2027 when traffic is down and AI won’t attribute, put exposure, customer behavior, and business outcomes on the same monthly report. If visibility rises but customers rarely mention AI, investigate whether your tracked prompts reflect how buyers actually research. If more customers report AI discovery, examine whether those customers convert and stay.

Humbly,

Kevin

As a premium subscriber, you get the 2027 SEO Planning Workbook…

Read more

Subscribe now

39% of US consumers have used AI for shopping, according to Adobe, and Salesforce says AI influenced 20% of purchases during last year’s BFCM. But only 11% let AI make purchases for them.

So AI’s influence sits in product recommendations, aka the shortlist. And the shortlist gets built from what your customers do after they buy: the reviews they write, the videos they post, the brand names they search. Here is how to build a compounding loop around that new behavior.

From clicks to shortlists

Growth Memo research has repeatedly shown that users accept and trust the shortlist AI serves them: 88% of participants in our AI Mode user behavior study took the shortlist as-is. Consequently, 64% of users clicked nothing on average, and only 23% visited an external site (directly, without clicking a link). Another user behavior study about shopping in ChatGPT shows 93% of sessions had no meaningful click-out; only 3% verified the answer on the open web.

AI Visibility strongly impacts the products users pick: In ChatGPT, AI share of voice and user choice correlate at 0.57 on average. The range runs from grocery (1.0) and supplements (0.9) down to fitness apps (-0.4) and coaching (-1.0). Being at the top of the shortlist is vital: 74% of participants in our study picked the top ranked product in AI Mode. Rank 3 and below get 10%.

The only exception: Shoppers pick a brand they recognize over rank 1. In our AI Mode study, shoppers named brand recognition as a trust signal in 34% of sessions, vs. 23% in classic search. IAB Australia found 59% of AI shoppers trust AI more when they’ve seen the brand advertised elsewhere.

That differs strongly from Search, where 56% of participants built their own shortlist (compared to 5% in AI Mode) and only 21% adopted the shortlist shown in AI Overviews. So, the more Google funnels users into AI Mode, the more important it is to appear in product shortlists.

Keep in mind Search is about product grids: 96% of shopping queries surface product grids in an analysis of 4K commercial queries. Shopping Search is not about blue links anymore. When AIOs appear, they cut clicks on blue links in half.

For AI shortlists, tactics you can control (like optimizing your product feed and onsite optimization) only get you so far. AI leans heavily on 3rd-party signals like video, web mentions, and reviews. AI visibility in Google’s AI Overviews has much stronger correlations with YouTube mentions 0.7, branded web mentions 0.7, branded anchors 0.6 and branded search volume 0.4 than metrics like DR 0.3 or referring domains 0.3, according to an ahrefs study. This citation pattern is confirmed in another analysis by Peec AI: Youtube is repeatedly cited as one of the strongest citation sources in AI Mode, AI Overviews, Gemini, and Perplexity.

The catch: none of these signals show up in a last-click report. A creator video that puts you on ChatGPT’s shortlist gets no credit when the shopper buys later. That’s why this episode’s sponsor, AirOps, runs a live session on investing in AI search before measurement catches up!

Join Matt Hammel, COO of AirOps, and Zak Prauer, Expert Partner, Marketing at Bain & Company, on Wednesday, October 15, at 3pm ET / 12pm PT. You’ll learn why every major channel mattered before anyone could measure it, why being recommended matters more than being cited, and how to keep pushing for better evidence without underinvesting.

Save your spot

Organic growth in ecommerce runs as a loop

I divide organic growth for ecommerce into 5 steps:

1. Being visible in the AI shortlist

2. Capturing more value out of the traffic we still get

3. Optimizing packaging for engagement

4. Investing in ongoing relationships

5. Driving brand trust and recall

At first glance, that’s a funnel with a few extra steps. But each stage hands its output to the next one. The shortlist earns the visit, the visit sells the product, first use builds the relationship, the relationship builds recall, and recall puts you back on the shortlist.

Stage 1: Optimize product feed to be eligible for the shortlist

The ecommerce ecosystem around Google, OpenAI, Shopify, and others is building the infrastructure of the future: shopping protocols. Product feeds mattered to earn appearances in Google’s product grids. Now, they’re critical for AI visibility and direct checkout.

Shopping protocols answer 3 questions for an agent:

1. What’s available?

2. Can I buy it?

3. What happened to the order?

Shopify found a 2x higher conversion rate for product listings in its catalog.

Do:

1. Connect feeds and checkout through UCP or ACP. Fill every feed field and use high-quality images.

2. Collect verified reviews on Trustpilot and your site.

3. Lead product titles with your differentiator. Descriptions say who it’s for, how it performs, and how it compares.

4. Answer 8 to 10 real buyer questions on each product and category page.

Stage 2: Maximize traffic you still get with tools

AI sends fewer, more qualified visitors, but they still need confidence on fit, taste, and comparison. Tools like quizzes, calculators and buyer guides help them decide and help you capture valuable 1st party data.

Quizzes can convert around 5% of takers into orders, roughly 2.75x the typical conversion rate. Beauty merchant Lawless built its quiz around the one question ChatGPT can’t settle for a shopper: which concealer shade fits. It converts at 36%, with 80% completion and a 34% higher AOV.

Quizzes also build a first-party email list. Octane AI, which sells quiz software, reports that its quizzes average a 29% email opt-in rate and capture 3x more emails than static signup forms. In its case study, Jones Road Beauty collected 800K email leads from its quiz, 80% of the brand’s 1M subscribers. Both numbers come from Octane AI’s own marketing, so read them as a best case. Still, there is something to it: a quiz asks for an email at the moment a shopper wants an answer.

Of course, that can also look “agentic”: Klaviyo’s support agent now resolves up to 65% of questions, providing real value to shoppers and merchants.

Do:

1. Pull 20 buyer questions from support tickets, reviews, and Reddit.

2. Turn them into a quiz, calculator, or finder on product pages and the homepage.

3. Send answers to your email platform as zero-party data and trigger flows from them.

4. Compare quiz completers’ conversion rate and AOV with sitewide. Target: 2x. Feed repeated bot questions back into product pages.

Stage 3: Packaging can lead to AI citations

25B people watched videos with “unboxing” on Youtube, and we know YouTube mentions correlate strongly with AI mentions (in AI Overviews). Opening the box a product is delivered in is the first real contact with it. With only 3% of users verifying AI answers online, this is where perception moves from the AI’s answer to the product.

Packaging signals value: the box sets expectations for quality, price and brand. Ryder’s study of 1,300 US shoppers shows 45% say a great unboxing makes them want to buy again, and 19% say gift-like packaging gets them to post photos or videos. So, packaging can lead to UGC like videos, posts, and reviews that AI uses in answers.

Do:

1. Design the reveal: Make opening and recycling obvious; add one branded moment worth filming.

2. Ship a 60-second setup video.

3. Ask for reviews 7 to 14 days after delivery.

4. Work with YouTube creators on unboxing and product-in-use videos.

5. Answer real questions in relevant subreddits with a named account that discloses the affiliation.

Stage 4: Win the next purchase between purchases

As winning new customers through organic search gets harder to attribute, the intuitive reaction is to focus on repeat sales. The math favors the second order. Across 12,000+ Gorgias merchants, repeat customers make up 21% of customers and generate 44% of revenue (source). Roughly one in five customers drives nearly half the business, so the first order should start a plan for the second.

Starbucks shows the mechanism at scale: today’s purchase builds toward a future benefit. Rewards members drive about 60% of Starbucks’ US revenue and are 5.6x more likely to visit daily (source).

It doesn’t take a points program, though. Quaker Marine, a nautical clothing brand, sent customers a simple check-in after delivery. Three-week repurchase rose from 9.7% to 11.2%, a 16% lift, without a discount.

Think about this: Shipping confirmation emails have the lowest buying intent (16% CTR) but get the most attention (62% open rate). Are you using them to educate customers or get them excited about the product?

The pattern behind these examples is a loop inside the loop:

1. Purchase: the order is placed.

2. Experience: delivery and first use.

3. Customer signal: usage, a need or a problem.

4. Timely response: a reorder prompt, service or a reward.

5. Repeat purchase: the next order.

Triggered emails are 5.3% of sends, 41% of email revenue and earn 18x revenue per recipient vs. campaigns, according to the Klaviyo 2026 benchmarks.

Do:

1. Build welcome, post-purchase, replenishment, and winback flows.

2. Split those flows by quiz answers. A field you collect and never use adds nothing.

3. Time reorder prompts to consumption. Half of second purchases happen within 30 days.

4. Track loyalty redemption, not enrollment. Active members spend 10% more, redeemers 25%.

5. Report human CSAT separately so the bot’s numbers can’t hide a decline.

Stage 5: Become the brand shoppers ask for

A brand name in the prompt changes the question. Ask ChatGPT for the best minimalist wallet, and it recommends GRAMS28. Ask for the best Ridge wallet for you, and the only question left is which Ridge. Brand searches are sure wins in AI Search, just like in regular search back in the day.

Ridge CMO: “Probably 70% of our sales come from direct, organic, or branded search. And people aren’t just typing in Ridge Wallet for fun. Something makes them want to search that.”

Brand recall is the cheat code that bypasses any shortlist. As the child of the 4 previous stages, recall is driven by 3 factors:

1. Repeated exposure (creators, ads, word of mouth)

2. Distinctive cues (name, design, message)

3. Good experiences (product, packaging, service)

The first 4 stages “feed” all 3. Showing up on AI shortlists and product grids repeats your name (stage 1). The quiz and the bot turn the visit into a good experience (stage 2). The box delivers the distinctive cue and the first-use moment (stage 3). Every check-in and triggered message adds another exposure (stage 4).

Recall comes first in the shopper’s journey, but it takes the longest to build, so it sits last in the loop. Those memories also leave a public footprint of videos, reviews and web mentions, which is what the Ahrefs correlations from earlier pick up.

Do:

1. Ask 5 models what you sell, what it costs, and what your return policy is. Fix what’s wrong at the source.

2. Use one category descriptor on your site and every third-party profile. Audit for drift.

3. Publish your own comparison pages before a competitor writes them.

4. Measure unaided recall quarterly: a post-purchase survey for customers, a broader category sample for the market.

Long-term value comes from what’s hard to copy

Every stage of the loop has a part competitors can buy and a part they have to earn.

Any competitor can match shopping protocols, crawler access, email and SMS triggers, a support bot, a quiz tool. That makes it table stakes.

The defensible part takes time and effort but compounds: reviews, creator relationships, community, first-party data and the recall that makes shoppers type your name into a prompt.

That split is the budget question for 2027. Fund the copyable half until you’re even with competitors, then put the rest into the half that compounds.

Subscribe now

It’s planning season! And like every year, most execs start by setting a traffic target that breaks down into revenue. But depending on what industry they work in, that doesn’t make sense anymore for most brands.

Last November, we wrote that SEO budgets buy capacity and that click-based ROI had decayed. A week later, we argued that leadership buys controlled learning when certainty is off the table. Both couldn’t be more relevant today.

So what changed for your 2027 planning? How much of the work now happens where no click, cookie, or referral can see it?

Plan 2027 around the signals that happen before the visit… or despite it.

1/ What you need to change for 2027 planning

The November 2025 memo made 2 calls: Most SEO costs are fixed (salaries, tools, infrastructure), so the budget question is which capabilities to fund. And the split should come from scenarios instead of last year’s traffic. For example, 40% of capacity goes to digital PR, 30% to technical SEO, 20% to content operations, and 10% to research, adjusted by priority and weak spots. We also advised pitching tests and experiments that were well planned, had firm boundaries, and kill switches if they weren’t producing results.

Keeping all that in mind, 3 things have moved since then:

• Zero-click searches rose from 60.45% of US Google searches in 2024 to 68.01% in early 2026, per SparkToro and Similarweb data. Roughly 2 in 3 searches now end without a visit, so a traffic baseline is doomed to shrink and undercount visibility before 2027 even starts.

• AI tools moved into marketing teams as unbudgeted build time. Part 1 of the AI marketing automation series argues those hours often come straight out of brand work.

• Attribution broke past the point of repair. Graphite estimates AI can be underattributed by 10x. Yikes. That means a plan developed on what can be cleanly attributed will cut resources from the channels that work.

2/ The visit isn’t where your planning should start

A buyer can read an AI answer, scan a Reddit thread, and see a peer’s LinkedIn post about your product before visiting your site once. In years past, planning for organic visibility treated the visit as the first measurable moment—or the ideal one. In 2027, most of the influence will land earlier.

Look at this search journey for buying a Samsung TV, measured in Semrush, for example.

Or this one, of a user buying a foundation from Sephora.

Disjointed. Messy. The Messy Middle!

And so far in 2026, only about 1/3 of searches produce any kind of click, according to Sparktoro. But the visits are still happening; they might just show up in the wrong column.

SparkToro and Similarweb tracked clickstream behavior for 7 days after AI recommended Capital One: Direct visits to those brands rose, while search visits ran about 15% lower, and the lift reached 14.2%. (So your dashboard that reads direct traffic as “unknown” or uncategorized is likely counting AI influence as noise.)

I chatted about this recently with my colleague and friend, Amanda Natividad, VP of Marketing at SparkToro and author of Zero Click Marketing, who puts the planning consequence plainly:

“If people are increasingly getting answers without clicking, then you need to plan for the signals that happen before, or instead of, the visit. That means prioritizing things like brand mentions, citations, discussions about your brand, and participation in the communities where your audience already hangs out.”

Those signals are spread thin across surfaces, with the AI citation/AI mention overlap between engines being minuscule: 91% of citations appear in only one of ChatGPT, Perplexity, or AI Overviews. So presence has to be planned surface by surface, much like social platforms.

Before you allocate 2027 hours to PR, content, or community, check which brands AI recommends in your category and which sources it cites.

Semrush is sponsoring this week’s memo, and its free AI Visibility Index lets you compare thousands of brands across 22 industries and four AI platforms. Its flagship study draws on 126M+ real AI search prompts.

Use the source data to shortlist places where your team could earn mentions next year, and the accompanying free studies and playbooks to choose tactics to test. Bring that shortlist into your planning meeting: which sources deserve a person’s time, and how many hours can you commit?

3/ Off-site distribution and community time need their own budget line

This year, if you gave a hefty amount of your planning resources to digital PR, it makes sense: 2025 and 2026 analyses found that it’s crucial to AI visibility. In fact, AirOps found that third-party signals drive 85% of brand visibility in LLMs.

Natividad extends that line past earned third-party signals:

“Budget for creating and distributing content beyond your own website, and actually engage with those communities. Then focus on creating ideas worth talking about, that also get mentioned and cited by other credible sources, and showing up in the places where people are already having relevant conversations.”

Community participation is labor, and should be planned for accordingly. And a community or UGC line with no named owner is the first one cut when Q2 gets tight, so budget it as hours per week with a person attached.

“Ideas worth talking about,” as Amanda mentioned, is a research cost. In 2027, fund the proprietary data that gives other sources a reason to cite you, the core of the brand authority series. Community expenses can be folded into the same capacity line item as SEO because its cost is pretty much fixed. Spend 5% (maybe 10% in some cases) of your marketing budget on community. (There’s a slide deck in the premium library to help explain this concept of brand authority multiplication to your team.)

4/ No single metric proves influence, so measure the signals as a set

Natividad’s measurement advice starts where attribution stops:

“Measurement has to expand, too. Look at search and AI visibility, citations and mentions, but also platform signals like impressions, engagement, branded search, and eventually conversions. None of those metrics individually proves that you influenced someone. But together, they give you a much better picture of whether your brand is actually reaching people.”

Last week’s memo on the collapse of attribution called this triangulation:

Combine three signals with different blind spots: an exposure metric, a behavioral signal, and a business outcome. For AI visibility, that could mean Share of Voice, self-reported discovery or CRM tags, and conversions; confidence rises when they move together, while divergence tells you where to investigate.

Natividad’s own recommended list maps onto 3 layers:

Report all 3 layers on the same page every month. A few examples:

• Exposure up, but branded search flat. Mentions aren’t landing with buyers. A distribution problem: move content and participation into the communities and channels your audience uses.

• Branded search up, pipeline flat. People went looking and didn’t convert. This is a conversion problem: Fix the pages they land on, the offer, or the handoff to sales.

• Branded search up, direct visits up, AI citations flat. Something off-site is working (a podcast, an event, a partner, word of mouth) and AI hasn’t caught up. Find the source and fund more like it.

• Exposure and branded search both down. Someone else is taking the mentions. This is a competitor problem: Audit who AI recommends in your category now and which sources it cites for them.

Each mix of the three layers points to a different fix.

“All of these systems are trying to understand and reflect human behavior,” Natividad shared with me. “Mentions, citations, discussions, engagement, and recommendations are all proxies for real people paying attention.”

And that’s what you need to communicate to the c-suite.

5/ Sell 2027 as a “planned portfolio” with kill dates.

Set your 2027 organic targets as presence on surfaces where your audience is and your potential buyers are researching. Develop goals to build:

1/ Visibility with AI engines your customers actually use

2/ Authority with online communities and publications where your category gets discussed, and

3/ Organic branded search volume.

Last year, we were all working to sell experiments to our leadership teams. Your 2027 plan, however, needs 3 separate “buckets” of work that you’re resourcing, each with its own rules:

• Proven work, funded at full capacity: The tactics (born out of 2026’s learnings) that are providing you competitive growth in your vertical, along with technical SEO, content refreshes, and the digital PR that already earns citations.

• Learning work, capped with one owner and a kill date: AI marketing team automation pilots and new search surfaces you haven’t tested.

• Unmeasurable bets, funded on judgment: community, events, and brand work tests. Ramp’s George Bonaci said in last week’s memo that attribution alone “would have led us to cancel all brand marketing, stunts, direct mail, events.”

Current search volatility argues for shorter commitments, so lock the annual budgeting total you can work from, and re-split the buckets every quarter against the 3-layer measurement set. The Growth Memo premium library has the SEO budget planning tool to help you build the case.

I would recommend quarterly re-forecasting, and many of the CMOs that I work with are comfortable with this regular reevaluation of resourcing. Ultimately, your team needs the freedom to adapt as things move before the next biannual or annual budget planning.

Directors who walk into 2027 planning without a defensible AI SEO investment case will lose budget to paid. But you don’t have to.

Read more

Read more

Read more

Is your team building something that will save time or cost time?

The AI Automation Work Router is now in the premium library.

Here’s what it does:

It takes one piece of AI automation work through 8 yes…

Read more

Read more

Exec Summary

Amazon closed the 21 days to August 31 at its lowest visibility since tracking began in November 2022: 1,705, down 16.9%, in a window where Google ran a spam update and served fewer PDFs for 2 weeks.

Google’s August 2026 spam update ran from August 18 to 21. The index slipped 2.4% across 1,842 companies in the 21 days to August 31.

• The IRS lost 59.6% (69 to 28), the Small Business Administration 52.8% and the Department of Education 30.8% in the 21 days to August 31, while Google served fewer PDF files in results. The other 40 federal domains in the index round-tripped.

• Ecommerce fell 7.9% across 290 companies in the same 21 days, with Home Depot -14.7% and Wayfair -26.5%, while Walmart (+5.9%) was the only large retailer up.

• YouTube gained 8.7% (3,391 to 3,687), its first up-window after 2 down issues, while Reddit fell 9.3% to its lowest level since April and TikTok fell 20.3% to its lowest since September 2025.

• Images rose 13.9% in the 21 days to August 31, the largest vertical gain, with the paid stock libraries up 23.8% together (Getty Images +24.4%, Shutterstock +22.9%) while the free libraries slipped 2.7%.

• Wikipedia lost 5.6% in the 21 days to August 31, from 8,280 to 7,820, the largest raw loss in the index at 460 points, and References fell 4.8% as a vertical, the largest share loss in the index.

Let’s dig in.

SEO landscape check

The index slipped 2.4% across 1,842 companies in the 21 days to August 31. Six trends carried the move.

1/ Amazon fell to a record low and Ecommerce gave back its August gain

Amazon lost 16.9% in the 21 days to August 31, from 2,051 to 1,705, and closed below every reading in its history, with data back to November 2022. The same 3 weeks cost Amazon 1.1% in 2023, 0.7% in 2024 and 9.5% in 2025. This year’s loss is nearly double the worst prior August, on top of a slide from 3,560 in November 2024 to 2,226 in January 2026, and it landed while the index around it slipped 2.4%.

The ad budgets moved in the same direction:

• Back-to-school spending on TikTok rose 20 to 40% year over year across July and August

• TikTok Shop sales grew 84% in the year to February and

• Go Fish Digital’s clients shifted money to TikTok after hitting “max spend on Amazon and Google,” where they found it “right behind Amazon” in sales volume.

Google search costs rose about 20% in August at the same time. Amazon’s organic shelf shrank in the weeks advertisers hit their ceiling on Amazon and Google and routed the overflow to TikTok Shop. TikTok itself sits at its lowest Google visibility since September 2025 (below), so the platform winning the retail budget is losing Google’s shelf while it does.

Ecommerce fell 7.9% across 290 companies in the 21 days to August 31 and gave back its August gain, with Amazon carrying 346 of the vertical’s 451 lost points. The rest of the category followed, all in the same 21 days:

• Home Depot -14.7% (315 to 268), after +27.0% in GIB 23

• Wayfair -26.5%

• Best Buy -9.4%

• Target -5.7%

• Walmart +5.9%, the only large retailer up

2/ The August spam update re-sorted the index for one week

The August 2026 spam update ran from August 18 to 21, inside calendar week 34 (August 17 to 23), and trackers logged more volatility after it ended with no announcement.

Week 34 moved 40 of 754 companies with a baseline above 3 by more than 20%, a 5.3% rate, half of week 31’s 11.0%. The weeks on either side were quiet: 2 companies in week 33, none in week 35.

Google stopped enforcing site reputation abuse manual actions in the EEA from August 30 under a European Commission mandate; flagged sections there get separated in ranking rather than suppressed.

One more Google change landed during the same weeks. Google now routes search clicks through google.com/goto redirects at what Nozzle’s Derek Perkins logged as a near-100% rollout after testing that began in July, aimed at scrapers. So now, any rank tracker has to follow the redirect to see the URL; check your tracker’s data continuity from late August before reading its numbers.

3/ Google served fewer PDFs and the IRS paid for it

Read more

Read more

Read more

In Monday’s Memo, I shared how 5.32 million search results showed Google demoting substitutable/replaceable listicles, not all listicles themselves.

As a premium subscriber, this week you get the Comm…

Read more

Read more

Exec Summary

Search volatility is spiking: 11% of companies in the SSI with real visibility moved more than 20% in calendar week 31 (July 27 to August 2), against 1.2% the week before, and the 4 most volatile weeks in the index’s history all landed in 2026. Google announced nothing.

Three weeks after GIB 22, 82 companies saw more than 20% volatility in a single week, compared to 9 companies the week before.

• ZipRecruiter gave back its entire July spike: down 31.2% in calendar week 31 and 24.6% for the 26-day window, closing at 76, below its June 15 level of 77.

• Jobs (-9.4%) and Local Search (-15.5%) flipped from leading the July window at +12.4% and +10.6%, while Delivery stayed positive at +10.0%.

• The dictionary cluster gained 10.7% in the 26 days to August 10, and References posted the largest share gain in the index (+1.40 points) for the 2nd consecutive window. The last 2 Augusts were flat.

• YouTube lost 17.3% in the 26 days to August 10 after 6 straight down weeks; Social Networks fell 5.9%, while Reddit sat still at +1.1%.

• Aggregators hold 76.0% of the SEO visibility split between the 2 company types, and only 38.3% of ChatGPT’s citation pool (August 4 fetch).

Let’s dig in.

SEO landscape check

Calendar week 31 delivered the 3rd sharpest weekly shuffle in the index’s history, with data back to November 2022.

• 82 of 747 companies with a baseline above 3 moved more than 20% in 7 days, an 11.0% rate against 1.2% the week before and 0.1% the week after.

• Only week 19 in May (15.4%) and week 15 in April (11.6%) rank higher, and the 4 sharpest weeks on record all landed in 2026.

Google confirmed no update. Trackers logged heavy volatility August 1 to 3 with no announcement (Search Engine Roundtable), and the index total barely registered it, slipping 0.5% across 1,834 matched companies. The shuffle ran under a flat surface.

July’s spike names gave it all back

ZipRecruiter, GIB 22‘s winner, hit an all-time high of 113 on July 27, then lost 31.2% in week 31. It closed the 26 days to August 10 at 76, slightly below its June 15 level of 77.

Its entire July gain is gone.

Yellow Pages shows the same pattern: An all-time high of 28 on July 27, a 28.3% drop in week 31, and a 13.4% loss for the window after 3 consecutive issues on the Movers list. The same reversal shows up at the vertical level, all in the same 26 days:

• Jobs fell -11% after leading the July window at +12.4%

• Local Search fell -16.7% after gaining 10.6%

• Delivery held at +12%, the only July leader still positive

The losses stayed concentrated in last month’s spike names. Across 564 companies with data in both windows, the correlation between July moves and August moves is -0.08, essentially 0, so the index as a whole didn’t swing back. Whatever re-sorted, week 31 hit July’s one-week spikes first.

Angi, GIB 22‘s loser, stopped falling: down 3.9% for this window to 2, with its first up-week since June 1.

Dictionaries collected the shelf space

The 7-site dictionary cluster gained 10.7% in the 26 days to August 10, from 2,528 to 2,798, and References posted the largest share gain of any vertical (+1.40 points, n=80) for the 2nd consecutive window (GIB 22: +4.6%).

• Merriam-Webster +9.4% to 1,412, 0.9% below its May 18 peak of 1,425

• Dictionary.com +21.3% to 439, still 37.9% below its September 2023 peak of 707

• Oxford Learner’s Dictionaries +37.8%

• Vocabulary.com +21.0%

• The Free Dictionary +58.3%, a top-5 Mover this window

This is not the usual back-to-school bump. The same July-to-August window was flat in the last 2 years (-0.8% in 2024, +1.6% in 2025) and rose 13.4% in 2023 for the sites tracked back then. A normal August doesn’t produce a 10.7% gain.

The catch: The Wikimedia side of the reference stack went the other way. Wiktionary fell 23.4%, and wikimedia.org fell 32.0% in the same 26 days, while en.wikipedia.org added 268 raw points, the largest raw gain in the index. Ecommerce collected shelf space too, gaining 9.5% across 291 companies, with Amazon up 12.8% and Home Depot up 27.0%.

YouTube extended its slide while Reddit sat out

YouTube lost 17.3% in the 26 days to August 10, from 4,100 to 3,391, its 2nd consecutive issue in decline after -12.2% in GIB 22. The slide ran 6 straight weeks from the June 22 peak of 5,288 before its first uptick on August 10, a 35.9% drawdown in 7 weeks. YouTube alone accounts for 97% of Entertainment’s 12.3% vertical decline.

The rest of social followed, all in the same 26 days:

• Facebook -9.2%

• Instagram -11.6%

• Pinterest -7.5%

• Snapchat -30.8%

Reddit sat this one out. After +18.1% in GIB 21 and -14.0% in GIB 22, it closed the window at +1.1%, at 2,338. Two windows of whipsaw, then stillness.

A related Google move landed in the same window: Search Console now tracks how Instagram, TikTok, X and YouTube content performs in Search and Discover through platform properties (Search Engine Roundtable, July 7), rolled out to all users since July 29. The platform domains are losing SERP visibility while their content becomes a search surface you can measure directly; verify your accounts and see which queries route to your profiles.

The bigger story sits on the AI side: ChatGPT cites a different set of companies than Google’s AI surfaces do. Let’s Double Click.

Behind the gate: The Top 5 Movers and Shakers, Winner and Loser breakdowns for The Kitchn and StreetEasy, and the Double Click on the brands ChatGPT cites that Google’s AI surfaces skip. Upgrade to read the full issue.

Subscribe now

Read more

Read more

This week, you get a new stakeholder slide deck. Module 1 makes the investment case to leadership. Module 2 trains your team o…

Read more

Read more