When building is cheap, distribution becomes the moat

AI collapsed the cost of making software and content. It did nothing to the supply of attention and trust, so that is where the fight moved.

Mumbai dabbawalas sorting rows of tiffin boxes for delivery
Photo: Steve Evans from Citizen of the World, CC BY 2.0, via Wikimedia Commons

A distribution strategy for startups now matters more than the product build, because AI has made building cheap while attention and trust stay scarce. Controlled studies show AI coding tools can speed up well-defined tasks by over half, so competitors appear faster and in greater numbers. Meanwhile Google sends far fewer clicks when an AI summary appears, and AI assistants send little traffic yet. The startups that win own a route to buyers: an audience, a community, a sales motion, or a place in the sources AI answers trust.

Key takeaways
  • 55.8% faster: what GitHub Copilot did to one well-defined coding task, so new competitors arrive sooner.
  • 19% slower: what early-2025 AI did to experienced developers on mature codebases, so incumbents keep their edge in customers, not code.
  • 8% vs 15%: Google users click a normal result far less often when an AI summary appears (Pew, 2025).
  • 0.1% of referral traffic came from AI assistants in 2025; small, but the sign-ups it sends are well qualified.
  • Owned channels (email, community, founder audience, brand) compound; search, social and ads are rented.

Why did AI make distribution the bottleneck?

AI made distribution the bottleneck by cutting the cost of supply without touching the supply of buyer attention. When it took a team six months to ship a credible product, the build itself filtered competitors.

When a competent founder can ship in weeks, the filter moves downstream, to who can get the product in front of the right people and be believed.

Same complaint, sharper every year

Three Ask HN threads tell the story better than any survey. In December 2024 a solo founder asked how to learn marketing and sales at all; it drew over 200 comments.

  • April 2026: another founder said every product followed the same arc: build, post, a dozen likes from friends, then nothing.
  • July 2026: a pair of SaaS founders asked how to market "in the age of slop", because launching had become so easy that every channel was saturated with new products.
55.8%faster coding on a set task with GitHub Copilot (2023 experiment)
8% vs 15%Google searches ending in a result click, with and without an AI summary (Pew, 2025)
58%lower position-one CTR on keywords with AI Overviews (Ahrefs, Dec 2025 data)
0.1%of website referral traffic from AI assistants (Ahrefs, 2025)
The story

On 4 July 1996, Sabeer Bhatia and Jack Smith launched Hotmail, a free web email service. The product worked, but sign-ups were modest: about 100 in the first hour and roughly 250 that day. Their investor, Tim Draper of Draper Fisher Jurvetson, kept pushing one idea: put a line at the foot of every outgoing email inviting the reader to get a free account. The founders resisted; it felt like polluting their users' mail. They finally agreed, dropping the "P.S. I love you" opener Draper had first suggested. Growth turned into a hockey stick, averaging about 3,000 new users a day. Hotmail had 750,000 users by Labor Day 1996 and a million within six months. The product had not changed. Its route to new users had.

TechCrunch: PS: I Love You. Get Your Free Email at Hotmail (book excerpt, Oct 2009)
Before cheap AI buildingAfter cheap AI buildingMany ideasMany ideasBuild: 6+ months (filter)Build: weeksFew products reach marketMany products reach marketDistribution: easierDistribution (filter)Winner: who built itWinner: who is trusted
The filter on competitors used to be the build. With AI, it sits at distribution: attention and trust.
The Impact of AI on Developer Productivity: Evidence from GitHub CopilotSida Peng, Eirini Kalliamvakou, Peter Cihon and Mert Demirer · arXiv 2302.06590 · 2023

Recruited developers were asked to build an HTTP server in JavaScript as fast as possible; the group with GitHub Copilot finished 55.8% faster than the control group. The authors also saw uneven effects across developers that suggest AI can help newcomers move into software work. The limit: one well-specified greenfield task, which is the easiest case for an AI pair programmer.

Read the paper (free)
Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer ProductivityJoel Becker, Nate Rush, Elizabeth Barnes and David Rein (METR) · arXiv 2507.09089 · 2025

A randomised trial with 16 experienced developers working on 246 real tasks in mature open-source projects they knew well, each task randomly allowed or barred from using AI tools such as Cursor with Claude 3.5 and 3.7 Sonnet. Developers expected AI to cut time by 24% and afterwards believed it had cut time by 20%. In fact, AI use increased completion time by 19%. The limit: a small sample on large, mature codebases, the hardest case for AI, with early-2025 tools.

Read the paper (free)

What the two studies mean for marketers

Read together, these two studies matter for marketers more than for engineers. AI makes it dramatically cheaper to build new, simple things, which is exactly what most startup products are at launch. It helps much less with large, mature systems.

So the number of new entrants in any software niche rises, while the incumbents' advantage shifts away from code and toward customers, data, brand and distribution. The edge is no longer "we built it"; it is "they already know and trust us".

StudyTaskResult with AIWhat it implies
Peng et al. (2023)New HTTP server, greenfield55.8% fasterNew products appear faster
Becker et al., METR (2025)246 tasks on mature codebases19% slowerIncumbents keep their code edge, so the fight moves to customers
A Mumbai dabbawala carrying tiffin boxes
Mumbai's dabbawalas are a distribution network, not a cooking business. The food is easy to make; getting it to the right desk every day is the hard part.Photo: Ninara, CC BY 2.0, via Wikimedia Commons

Is search still a reliable way to get users?

Search still sends most referral traffic, but each ranking now earns fewer clicks, especially on informational queries. Pew Research tracked 900 US adults' real Google searches in March 2025.

When an AI summary appeared, users clicked a traditional result in 8% of visits, against 15% when there was none. They clicked a link inside the summary only 1% of the time, and were more likely to end their browsing session.

What Google users did, with and without an AI summary on the page
See the data as a table
Page with AI summary (%)Page without AI summary (%)
Clicked a traditional result815
Ended the browsing session2616

Source: Pew Research Center, 900 US adults, 68,879 Google searches, March 2025

The click loss is getting worse

Ahrefs' data on 300,000 keywords points the same way and is getting worse with time. Its April 2025 study estimated that AI Overviews cut the click-through rate of the top-ranking page by 34.5%.

Its update using December 2025 data put the drop at 58%: position-one CTR on keywords with AI Overviews fell from 0.073 to 0.016, while informational keywords overall fell from 0.076 to 0.039.

The replacement channel is still small

In Ahrefs' March 2025 analysis of about 35,000 websites, AI assistants sent roughly 0.1% of referral traffic, against 43.8% from search, 42.3% direct and 13.1% social. Google still sent about 345 times more traffic than ChatGPT, Perplexity and Gemini combined.

Yet Ahrefs said about 3% of its own sign-ups over the previous year came from AI, which suggests AI referrals are small but well-qualified.

Where website referral traffic came from, about 35,000 sites
See the data as a table
Share of referral traffic (%)
Search43.8
Direct42.3
Social13.1
Paid0.5
Email0.2
AI assistants0.1

Source: Ahrefs AI traffic research (Mar 2025)

How do you get your startup into AI answers?

You get into AI answers by being in the sources those systems retrieve and by writing content that is easy to quote. The first academic study to name this field found that specific content changes can raise a source's visibility in generative answers by up to 40%.

GEO: Generative Engine OptimizationPranjal Aggarwal, Vishvak Murahari, Tanmay Rajpurohit, Ashwin Kalyan et al. · arXiv 2311.09735 · 2023

The authors built GEO-bench, a benchmark of diverse queries across domains with the web sources needed to answer them, and tested how edits to a source change how often and how prominently generative engines cite it. Well-chosen methods, such as adding statistics, quotations and citations, raised visibility by up to 40%, and what worked varied by domain. The limit: tests ran on research versions of generative engines in 2023, and commercial systems change their retrieval often.

Read the paper (free)

In practice that means three things:

  1. Publish quotable answers. Specific, sourced answers to the questions your buyers ask.
  2. Be mentioned where AI already looks. Forums, reviews, comparison sites and Wikipedia-grade references.
  3. Make your own site easy to crawl and parse.

I cover the mechanics in how AI Overviews and ChatGPT choose sources and the step-by-step in how to get cited by ChatGPT and Perplexity.

Which distribution channels can a startup own?

A startup can own an email list, a community, a founder's personal audience, direct relationships with customers and partners, and its brand name in people's heads. Everything else is rented: search rankings, social reach, marketplace placement and ad auctions can be repriced or switched off by someone else.

Owned: compoundsRented: can be repricedEmaillistCommunityFounderaudienceGoogleorganicAIanswersSocialreachPaidadsGoal: move people from rented reach into owned lists
Every rented channel should feed an owned one: a follower becomes a subscriber or a community member.
ChannelOwned or rentedMain riskWhat compounds
Email list and WhatsApp opt-insOwnedList fatigue, deliverabilityDirect reach that no algorithm controls
Community (forum, Discord, events)Mostly ownedNeeds constant careMembers recruit and answer for you
Founder's LinkedIn or X audienceHalf-ownedPlatform reach changesPersonal trust that transfers to the product
Google organic searchRentedAI summaries cut clicksAuthority, if content is genuinely useful
AI assistant answersRentedOpaque source selectionMentions across trusted third-party sources
Instagram and YouTube reachRentedAlgorithm shiftsFollowers who convert into owned lists
Paid adsRentedRising auction pricesLittle, unless the brand is building

What the founders on Hacker News said worked

The HN answers that held up were mostly about owned channels.

  • On the April 2026 thread, a solo founder who said he had crossed $100K in monthly recurring revenue described talking to every customer he acquired.
  • Others argued that direct outreach to people in the forums and chats where the problem is discussed beats broadcasting.
  • On the July 2026 thread, one reply observed that paid search was producing clicks with dismal conversions while forum engagement converted well, and suggested answering questions in two or three forums where buyers actually ask them.
Crowded lanes and shopfronts in the Chandni Chowk area of Old Delhi
Chandni Chowk works because buyers already go there. Communities are the digital version of a market street.Photo: Pecha tenzin, CC0, via Wikimedia Commons

How should a technical founder approach distribution?

A technical founder should treat distribution as part of the product from the first week, not as a launch task. The most common failure in those HN threads is the build it and they will come arc: months of building, one launch post, silence.

  1. Find the room before the product. List the forums, groups, newsletters and events where your buyers already discuss the problem. If you cannot find one, that is a warning about the market.
  2. Talk to people one by one. Direct outreach does not scale, which is why it works early: it buys learning and the first ten customers.
  3. Publish what you learn. Specific posts with numbers and real examples become founder audience, search content and AI-citable sources at once. My founder-led LinkedIn content system shows a weekly rhythm.
  4. Turn every channel into an owned list. A follower, a reader or an event attendee should be invited to email or a community.
  5. Show up in person where deals happen. For B2B, events still concentrate buyers; my post on B2B event ROI covers when they pay back.
  6. Add paid only to amplify what already converts. Ads multiply a working message; they rarely find one.

Events are distribution you can stand in

I have seen this from the B2B side. At Gnani.ai, a voice AI company, I worked in marketing through a period that saw more than ₹40 crore of enterprise pipeline and eight-plus Indian and international events in about 18 months, among them a 221 m² showcase at the India AI Impact Summit in February 2026. A booth is distribution you can stand in. More of that record is on my work page.

Does distribution beat product quality?

Distribution does not beat product quality; it decides whether quality gets a chance to be noticed. A weak product with great distribution churns its customers and buys them again at a rising price. A great product without distribution simply goes unseen.

The change AI brought is that the second failure is now far more common, because many more good-enough products exist.

ProductDistributionWhat happens
WeakStrongChurns customers, rebuys them at rising cost
StrongWeakGoes unseen; now the common failure
StrongStrongCompounds: trust, referrals, citations
Code is easier to copy than ever. An audience that trusts you is not.Sapta, this post

That is why distribution is becoming the moat. Code is easier to copy than ever. An audience that trusts you, a community that answers your support questions, a brand name people search for, and a place in the sources AI systems cite, all take years to build and cannot be cloned by a weekend of prompting.

If you are building that for a company, my hire page says how I can help. For the specific channel mechanics, start with how the Instagram Reels algorithm ranks content.

Questions people ask

How do you handle marketing as a solo technical founder?

Start with direct conversations in the places your buyers already gather, publish what you learn in specific, useful posts, and turn every reader into an email or community member. Treat these as weekly product work, not a launch event. Paid ads come later, once a message converts.

What is the best distribution channel for a new startup?

The one where your buyers already ask about the problem. For developer tools that is often communities and GitHub; for B2B it is often LinkedIn, events and direct outreach; for consumer products it is short video and creators. Pick one or two and go deep rather than spreading thin.

Is SEO still worth it in 2026?

Yes, but differently. Clicks per ranking are falling where AI summaries appear, so pure traffic plays are weaker. Content that is specific, sourced and quotable still earns rankings, AI citations and trust, which together are worth more than traffic alone.

How do I market in the age of AI slop?

Compete on what slop lacks: specific experience, real data, named people and genuine community presence. Answer real questions where buyers ask them, and keep your own audience so you are not dependent on any one algorithm.

Sources

  1. Peng et al.: The Impact of AI on Developer Productivity: Evidence from GitHub Copilot (2023)
  2. Becker et al. (METR): Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity (2025)
  3. Aggarwal et al.: GEO: Generative Engine Optimization (2023)
  4. Pew Research Center: Google users are less likely to click on links when an AI summary appears (2025)
  5. Ahrefs: AI traffic research (2025)
  6. PPC Land: Google's AI summaries now swallow 58% of clicks (2026)
  7. Hacker News: Ask HN: How do you do marketing in the age of slop? (2026)
  8. Hacker News: Ask HN: How do you handle marketing as a solo technical founder? (2026)
  9. Hacker News: Ask HN: How to learn marketing and sales as a solo entrepreneur? (2024)
  10. TechCrunch: PS: I Love You. Get Your Free Email at Hotmail (excerpt from Adam Penenberg's Viral Loop) (2009)
distribution strategy startupsgo-to-marketowned audiencecommunity marketingAI searchGEOfounder-led marketing
Sapta

Sapta (Saptarshi Sarkar) is a mechanical engineer turned marketer in Bengaluru and a solo touring rider. He writes the seven Saptaverse blogs. Work with him or read more of Marketblog.

Keep reading in Marketblog

From the other blogs