ChatGPT sends 0.62% of Google’s traffic. Now what?

ChatGPT sends roughly 0.62% of the organic traffic Google sends, and the visits it does deliver appear to be worth less per click than Google’s. So the honest answer to “now what” is: give AI search a measurement line and a content policy, not a budget reallocation. The more consequential news this week sat on the supply side — how much of your content AI systems are already consuming, and how little say you had in it.

Four threads from the week that actually change something on Monday morning.

1. AI referrals are real, small, and cheaper than they look

Search Engine Journal gathered four separate datasets on AI visibility and landed on a figure worth writing down: ChatGPT accounts for about 0.62% of the organic traffic Google drives. Not zero. Also not a channel shift.

MarTech supplied the commercial half of that picture, reporting that ChatGPT’s paid traffic is worth far less than Google’s — thinner intent, weaker conversion, identical cost to serve.

The tooling market, meanwhile, is sprinting ahead of the traffic. HubSpot published a head-to-head of Scrunch and Ahrefs for AI visibility tracking. That comparison existing at all tells you where we are: vendors are selling dashboards for a channel most brands have not yet sized.

Takeaway: Track AI visibility as share of answers, not share of sessions, and keep it on a separate line from search engine optimization (SEO) reporting. Before you buy a tool, filter your own analytics for referrals from chatgpt.com, perplexity.ai and gemini.google.com. If that number is a rounding error, your problem is not measurement — it is that you have no reason to be cited yet.

2. Your content is already in datasets you never licensed

Brave sold access to its search index to AI companies, and as Search Engine Journal’s coverage framed it, your website is probably in it. Brave built that index with its own crawler — one almost nobody thought to block, because almost nobody treated Brave as an AI supply chain.

In the same week, SEJ’s SEO Pulse covered a Google AI-training opt-out control, a Lighthouse scoring update that will move PageSpeed numbers without anyone touching code, and ongoing rank-data reliability problems across third-party tools.

Stack those together and the lesson is uncomfortable. Blocking Googlebot and GPTBot in robots.txt is the easy part. Your content reaches models through intermediaries — independent indexes, data resellers, crawlers whose names are not on anyone’s blocklist — and each hop strips away your ability to say no.

Takeaway: Pull a server-log audit of user agents this quarter, not just a robots.txt review. Then make one company-level decision: is your content a lead magnet that should spread everywhere, or an asset you expect to be paid for? Most teams land in the middle — let the educational 90% travel freely, gate the proprietary research, pricing data and benchmarks behind a form or a login. What you cannot do any more is leave that decision to the default.

3. Search demand is leaking out of search engines

Two stories, one direction. SEJ reported that TikTok search ads now look like real search advertising — keyword-driven, intent-led, familiar to anyone who has run Google Ads — with four meaningful catches in how targeting and reporting differ. Social Media Today separately covered TikTok launching a standalone shopping app, which is the same move in retail clothing: pull purchase intent into an owned surface.

Google is defending from the other side. Its Gemini Business Profile tools are coming to Workspace, so updating a local listing becomes something you do inside Gmail or Docs rather than a separate dashboard visit.

For Indian marketers, the TikTok half is mostly a skills story, not a media-buying one — the app has been unavailable in India since the 2020 ban, so this matters if you run US, Gulf or Southeast Asia campaigns. The Google Business Profile half matters immediately: for a huge share of Indian local businesses, the Business Profile is the storefront, and lower-friction editing means your competitors will update theirs more often.

Takeaway: Stop treating “search” as one channel with one team. Audit where your category’s questions are actually typed — a marketplace search bar, a short-video app, an AI assistant — and staff keyword work for the top two surfaces, not just Google.

4. Buyers did not stop. They got slower and warier

MarTech’s read on the demand side is that consumers are still spending, they are just “harder to convince”. That single shift explains a lot of this week’s other reporting.

It explains why MarTech asked whether your affiliate program is showing up too late — partners that only appear at the final comparison step are claiming credit for persuasion they did not do. It explains why Marketing Dive’s look at Lands’ End’s holiday pop-up is interesting: the physical activation is a data and proof engine feeding an always-on program, not a seasonal spike.

And it explains why consent is becoming a product problem. MarTech argued for engineering consent into the experience rather than bolting on an opt-in, while Social Media Today reported Meta, TikTok and Google pushing back on a California law. India’s Digital Personal Data Protection (DPDP) Act points the same way, with its consent-manager model expecting granular, revocable permission rather than one blanket checkbox.

Takeaway: Move persuasion earlier. Audit your three longest-converting journeys and ask what proof a skeptical buyer needs at step two, not step nine — reviews, comparison content, a named case study, a visible return policy.

The thread connecting all four

Distribution is fragmenting and getting cheaper per unit, while trust is consolidating and getting more expensive. AI assistants will send you a trickle of low-intent clicks while ingesting your entire library. TikTok and Google both want the search moment to happen inside their walls. Buyers want more evidence before they commit.

The assets that hold value across all of that are the ones you own outright: a site worth citing, consented first-party data, and proof a stranger believes. Everything else is rented.

What this means for you

• Add one row, not one budget line. Report AI referrals monthly from your own analytics. Revisit the spend question when they clear 2-3% of organic sessions.
• Audit crawlers before you audit robots.txt. You cannot block what you have not logged. Do this once a quarter.
• Write one content-licensing stance. One page: what spreads freely, what is gated, who approves exceptions. It will save you six arguments next year.
• Treat PageSpeed score changes as noise until verified. Lighthouse scoring updates move numbers without any code change — check your field data before anyone panics in a leadership review.
• Map your category’s second search surface. Marketplace, app store, short video or AI assistant. Pick one and run a 60-day keyword test.
• Rebuild one consent flow for granularity. Separate transactional, product and marketing permissions. Under DPDP-style rules this is compliance; in practice it also improves deliverability.
• Shift affiliate and partner payouts earlier in the journey. Reward introduction, not just the last click before checkout.

Frequently asked questions

If AI search is under 1% of Google’s traffic, can I ignore it?

No, but you should not fund it like a channel yet. Search Engine Journal’s compiled datasets put ChatGPT at roughly 0.62% of Google’s organic traffic, and MarTech reports that AI-sourced paid traffic converts worse. Measure it, make your content easy to cite and quote, and reassess quarterly.

Can I stop AI companies from training on my website?

Only partially. You can block known crawlers in robots.txt and use platform-level opt-outs where they exist, but content also reaches models through third-party search indexes — Brave’s sale of its index to AI companies is this week’s example. Assume anything publicly readable will eventually be ingested, and gate what you genuinely need to protect.

Are TikTok search ads relevant for marketers in India?

Not for domestic campaigns — TikTok has been unavailable in India since 2020. They matter if you buy media for US, Gulf or Southeast Asian markets, and the underlying skill transfers: keyword intent research now applies inside social apps, including Instagram and YouTube search in India.

What is the single best AI visibility metric to report?

Citation share: the percentage of your priority questions where an AI assistant names or links your brand. It is closer to the actual mechanic than traffic volume, and it moves before referrals do.