All Articles
September 3, 2026
Profound, a company that sells AI search visibility tracking, raised $180 million on 15 September 2026 at a $1.8 billion valuation. Six days earlier it published its own research showing that the number it sells does not have a single value.
That gap is the useful part. Tracking how AI assistants describe your brand is now a funded software category, and the best research inside it says answer engines recommend different brands to different people. Both things are true at once, and together they change what you should ask a vendor before you pay one.
Profound’s Series D was led by Sequoia Capital and Kleiner Perkins, with Lightspeed, Khosla Ventures and South Park Commons taking part. The company says revenue tripled over the previous six months and it now serves more than 1,000 enterprise customers, Comcast and Walmart among them, according to TechCrunch.
The timing is the detail worth keeping. This round landed less than seven months after a $96 million Series C, which means the category is being priced on growth rate rather than on proven attribution. Forrester now publishes an Answer Engine Optimization Technologies Landscape, so answer engine optimization has gone from a tactic to a vendor field with a buying process attached.
None of that tells you whether it works on your site. A funding round measures investor confidence in a market, not the return on a subscription for a business your size. Treat the round as evidence that the problem is real and widely felt, and nothing more than that.
Between 11 and 24 August 2026, Profound ran 71,147 responses through ChatGPT, Claude and Gemini, asking about clothing, furniture and credit cards while varying the persona by gender, age, income and occupation. The recommendations moved.
Income moved them most. High-income personas received 7.5 brand mentions on average; low-income personas received 5.8. Asked for a t-shirt, the engines returned $8 to $24 for low-income personas, $10 to $58 for middle-income ones, and $140 to $1,650 for high-income ones. Age and gender changed which brands appeared: Patagonia and Ralph Lauren surfaced for men, Reformation and Free People for women.
Here is the limit. The study covers three consumer categories over a fourteen-day window, and Profound could read back recorded search queries for ChatGPT and Claude but not for Gemini. It does not prove your category behaves the same way. It does show that “where do we rank in ChatGPT” is the wrong shape of question, because a single ranking is an average across personas, and the average persona may not be your buyer.
Google added generative AI performance reports to Search Console in June 2026. They cover AI Overviews and AI Mode, and they report impressions: how many times links to your site were shown inside a generative AI feature.
Impressions is all you get. Google’s own documentation confirms there is no click metric in the report and no query dimension. You can break impressions down by page, country, device and date, and that is the complete list. The standard 1,000-row limit applies, Search Labs experiments are excluded because they are still in development, and the newest days are marked as preliminary.
There is a way to see AI Mode queries, and it is not in the AI report. John Mueller of Google explained in August that “if a user asks a follow-up question within AI Mode, they are essentially performing a new query”, and that impression, position and click data from the response counts against that new query. Those queries land in the ordinary performance report, as Search Engine Roundtable reported on 6 August 2026. Filter last month for unusually short or conversational strings, the ones that read like half of a spoken exchange, and you are looking at AI Mode traffic with clicks attached.
This takes about an hour and costs nothing. Do it before you take a vendor demo, so you have your own baseline to argue from.
Step three is the one people skip and the one that pays. Running it yourself is also how you find out whether your brand authority survives the moment a model has to pick three names out of a category.
For most solo marketers and small teams, not yet. The tooling is priced for enterprises with 1,000-customer logos, and the free checks above surface the same directional signal. The case for paying gets stronger when you sell in a category where a wrong recommendation costs a lot, or when someone internally needs a number reported monthly.
Partly. AI Mode queries show up in the ordinary Search Console performance report with clicks attached, because follow-ups are counted as new queries. ChatGPT, Claude and Gemini referrals appear in your analytics as referral traffic, but without the prompt that produced them.
No. The persona research points at a narrower fix: be explicit about who a product is for and at what price, because that is the information the engines are sorting on. That is a change to a handful of pages, not to your whole content operation.
Stop asking where you rank in AI search and start asking who you rank for. Take the free baseline this week, run the three-persona test in your own category, and hold any vendor conversation until you can compare their number against something you measured yourself.
This is the kind of thing we work on with clients at Social Lady. If you want a second opinion on how your brand shows up in AI answers, get in touch.
