Founder, Genmark AI | AI Search, Growth, and Agentic Marketing | Ex Microsoft (Search Ads), AI/HR Tech Startup | Building the next growth stack
Bellevue, WA
Joined February 2009
- Tweets501
- Following1.3K
- Followers229
- Likes4
Pinned Tweet
Buyers now ask ChatGPT, Gemini and Perplexity which vendors belong on their shortlist. Most B2B teams have never checked what those answers say about them.
Here's the full playbook to check it and fix it, in 7 steps. No tools needed.
Bookmark this. One step per post below.
2
1
132
A mention that quotes last year's pricing still counts as visibility. It can be worse than no mention.
Keep a facts register: each fact, the correct value, one canonical URL, an owner. Check AI answers against it on a schedule.
1
9
10 questions for a win/loss interview. Ask 5 recent buyers and 5 who chose someone else, within a month of the decision.
1/ What was happening that made you start looking?
2/ What did you try before looking for a tool?
3/ Where did you first hear about us?
4/ Which other options did you seriously consider?
5/ Did you ask an AI assistant for options? What did it say?
6/ What almost stopped you from choosing us? (Or: what made you choose them?)
7/ Who else had to agree, and what did they care about?
8/ What did our website or sales process get wrong?
9/ How did the price compare with what you expected?
10/ What would you tell a friend who's looking now?
How to use it:
-> Record the calls (with permission) and copy the exact phrases buyers use.
-> Put those phrases on your homepage, ads and pricing page.
-> Count the answers to questions 3 and 5. That's your real channel mix.
Buyers will tell you your positioning if you ask.
15
Dave's missive to struggling founders is worth reading in full. Underneath it is one question: how badly do you still want the job?
Three honest tests I'd add:
1. Which part of the week do you dread on Sunday night?
2. If you could hire one person tomorrow, would they take that part away, or take your seat?
3. Do you still have ideas for the company you haven't tried yet?
The answers point to an advisor, an operator, or the exit.
A Missive to Struggling Founders kellblog.com/a-missive-to-st…
1
50
Jason's point is the one SaaS pricing teams should sit with.
Seat pricing quietly relies on people not auditing their tools every day. Agents do. An agent sees the cost of every call, and it switches the day a cheaper tool does the job.
If you sell software into an agent's workflow:
1. Price on outcomes the agent can measure, not seats.
2. Make the API the product, not the upsell.
3. Assume your renewal happens every day, not every year.
Many folks (not just HubSpot) are working to refound their B2B companies now in the Age of AI. Good.
I’m going to give you one core bit of advice many won’t take but please at least consider it:
-> Make Your Product Cheaper. Potentially, Much Much Cheaper.
Why?
Then agents will use it more. Agents measure and track everything. Humans buy a product and sort of … forget. Not entirely, but partially, until we get a shocking bill and yet another price increase for the renewal cycle.
Not agents. If your product is too expensive, they stop using you as much. If it’s not cost effective, they seek out another solution. Right now. That day.
We’ve ditching apps that our agents couldn’t use as often as they needed to. We ditched Marketo, and we’ve moving off even of our own core AI products now, because they just make it too expensive per API call, per usage, etc.
Agents will use your product far more than humans, if you do it right.
So you have to make it cheaper.
Many, many will resist.
1
74
"A startup is a company designed to grow fast." That's the first line of Paul Graham's essay "Startup = Growth".
He says you can use growth like a compass for almost every decision. Most marketing teams don't, because they track too many numbers to steer by any of them.
Pick one:
1. Choose the single number that best predicts revenue for you: qualified demos, trials that activate, pipeline created.
2. Measure its weekly growth rate, not the total.
3. Before any new campaign, ask: will this move that rate? If you can't say how, it waits.
One number, measured weekly, makes the priority calls for you.
1
14
Organic traffic is down and someone on your team is already saying "AI killed SEO."
That's a guess, not a diagnosis. Here's the exact check I'd run in Search Console and GA4 before moving a dollar of budget: (1/8)
2
2
30
Step 5, price the loss. Tag lost pages by intent: how-to, comparison, pricing, demo. Multiply lost clicks by each group's conversion rate in GA4.
While you're there, check chatgpt.com and perplexity.ai referrals. Some of that demand moved, not vanished. (7/8)
1
6
If you run this and still can't tell which one is hitting you, DM me. Or grab 30 minutes and we'll look at your own data and competitors together:
genmark.ai/book/ai-search-gr…
Which step would surprise your team most? (8/8)
4
AI agent traffic to websites grew about 8x in two months, in one security firm's data. Almost 9 in 10 of those requests came from just two companies.
HUMAN Security's October "State of Agentic Traffic" report (sites and apps it protects, not the whole web):
- Meta's and OpenAI's agents made nearly 90% of agent requests.
- Meta's Muse, launched Sept 8, made 59% on its own.
- Total agent requests were about 8x August and 2.6x September.
- 85% went to product and search pages. Checkout and payment pages got 1.9%.
- Perplexity's Comet fell from 41% of requests in August to 3.6%.
An AI agent is software that browses for a person. It opens pages, searches and compares, like a personal shopper walking the store for someone. HUMAN's report also says most sites' analytics can't tell these visits from human ones.
Why it matters for B2B: agents are reading the same product and pricing pages your buyers read. If your bot protection quietly blocks them, or your pages only make sense to a human clicking around, you can lose the comparison before anyone talks to sales.
The same week, United Airlines put its lines in writing. Its terms, updated Nov 11: search engines may index public pages, any AI agent that searches, books or buys needs written permission, and its data may not be used to train or "ground" an AI model (grounding = an AI pulling in outside content to build an answer).
What to do this week:
1. Ask whoever runs your CDN or bot protection what happens today to Meta's Muse and OpenAI's agents: allowed, challenged or blocked. Make it a decision, not a default.
2. If you verify signed agents, update for Muse. It moved its signing keys from meta.com to muse.ai between Nov 7 and 9, and DataDome counted about half its requests unsigned on Nov 10.
3. Write your own three lines, like United did: search indexing, agents acting for a person, and use of your content to train or ground AI. Each can get a different answer.
4. Read your product and pricing pages as plain text. That's where 85% of agent requests landed.
One caveat: HUMAN and DataDome both sell bot detection, so treat these as one vendor's view of its own network.
Send this to whoever owns your website security settings.
Source and details below.
Follow @vinayah if you're working out how AI answers change B2B growth.
2
51
One more detail from HUMAN's report: ChatGPT Agent sent 44% more requests in October than in September, yet its share fell from 28.5% to 15.9% because Muse grew so much faster. Muse now sends 3.7x as many requests as ChatGPT Agent.
Report coverage (Search Engine Journal): searchenginejournal.com/meta…
United's new AI agent rules: searchenginejournal.com/unit…
18
Kevin Indig’s Similarweb-based analysis estimated organic visits down 13.2% for sampled US sites and 24.4% in France over two years.
But US decline was 7.8% excluding publishers. “AI killed SEO” is not a diagnosis. Segment first. #SEO
ALT Genmark AI infographic: Is SEO really dying? Growth Memo / Similarweb analysis by Kevin Indig, Oct 26, 2026, modeled traffic estimates for sampled sites, Aug 2024 to Aug 2026: estimated US organic visits down 13.2% and France down 24.4%; 62.5% of US sites and 66.6% of French sites in the dataset declined; publishers were hit hardest and smaller sites were more resilient; AI referrals were a small share of visits.
8
Earlier this week, I was at #Ignition2026 Summit organized by HubSpot for Startups.
Here are the key takeaways from the sessions (from my handwritten notes) and my conversations with other founders:
1. Narrow beats broad. One lane, one persona, one pitch, one package. And a first customer you treat as if they were your only one.
2. Distribution is the moat. With frontier labs moving into every category, the hard thing to copy is your ability to tell your story and get into the right rooms.
3. The founder is the megaphone. Recruit your first creators yourself, post every day, and invest in your personal brand sooner than feels comfortable.
4. People buy on emotion and justify it with logic. Sell by teaching, not pitching.
5. Make your first customer a champion. Get them something measurable, ask what value they got, then ask who else would be a fit while they're excited.
6. Customer success is product. Invest in CS early. Watch what repeats across customers, then build it into the product.
7. Hire your first seller sooner. Founder-led sales all the way to $1–2M is a myth, and the first hire doesn't need to be an AE.
8. Standardize before you delegate. Make the process repeatable first, then hand it to a person or an agent.
9. Context is the new layer. One shared layer across CRM, calls, usage, billing and support that every agent reads from. Agents judge the signals; a person still sends.
Thank you to the speakers: Yamini Rangan Grant Lee, Marcel Santilli, Michelle Lim, Nick Masters, Nico Chinot, Caroline Clark, Lisa Duncan, Evelyn Xu, Siqi Chen, Emily Bennett, Clayton Petty, CC Gong, Cristina Cordova, Gorish Aggarwal, Sriharsha Guduguntla
The best part of the day was the room. It was a highly curated group of founders, investors, and marketing Leaders, the conversations were as good as the stage.
Thank you to everyone I got a chance to talk to. I learned a lot from each of you. Thanks Maximus Greenwald for the clinic session chat.
Thank you to the HubSpot For Startups team for putting together such a thoughtful day. and to Dharmesh Shah (@dharmesh) who posted about the event opportunity on his LinkedIn :)
#Ignition2026 #HubSpotForStartups #GTM #Startups
Your Search Console clicks will never match your analytics visits. Google just published a page explaining why.
It's a new Search Central help page (updated Nov 12) on finding Google Search traffic in your analytics and server logs.
What it says, in plain words:
- Google only passes its domain as the referrer (google.com, google.co.uk). No page path, no search query. Queries live only in Search Console.
- Discover and Google Images send the same values, so a google.com referrer can't tell you which one sent the visit.
- Some Google visits arrive with no referrer at all: private browsing, new tabs in some browsers, jumping from a phone's search widget to a browser. In analytics, those can look like "Direct" visits.
- Search Console counts a click the moment someone taps. Analytics counts a visit only after your page loads and the tracking script runs. Ad blockers and consent choices stop that.
Why it matters: when leadership puts the two numbers side by side, the gap is not automatically lost traffic.
What to check this week:
1. Custom channel rules and referral exclusions. Google says they can push Search visits into Referral or Direct.
2. Redirects on landing pages. JavaScript redirects can replace the referrer, and redirects that drop the query string lose gclid.
3. If you count from server logs, count landing pages only, not image, CSS or JS requests.
One thing the page doesn't mention: AI Mode or AI Overviews.
Source and details below.
Follow @vinayah if you're working out how AI answers change B2B growth.
5
72
One more line from Google's page worth sharing with your analytics team: Search Console reports only organic clicks, while your analytics totals may include Google Ads traffic, depending on how it's set up. Compare like with like.
Google's page: developers.google.com/search…
26