{"slug": "bidding-on-competitor-names-when-buyers-ask-agents", "title": "Bidding on competitor names when buyers ask agents", "summary": "Steel, a browser-infrastructure startup for AI agents, found that competitors bid on its brand name in Google Ads while AI agents evaluating the product never see those ads, according to a first-person account from the company's growth lead. Raw server logs over two separate 24-hour windows showed Claude Code fetched Steel's site more than 10 times as often as Googlebot and made 4 to 5 times more requests than ChatGPT, while the company's analytics tool attributed only about 3% of sessions to AI assistants. Steel's clicks from searches for its own name grew about six times over the past year, with a click rate between 33% and 42% every month since December.", "body_md": "[← Writing](https://www.nibzard.com/log)\n\n# Bidding on competitor names when buyers ask agents\n\nRivals bid on our brand on Google while developers send AI agents to evaluate us. Agents never see the ads.\n\n## Share and tools\n\nLast week I searched Google for our own product. I typed “steel browser.”\n\nThree ads loaded before our name. One of them had our name in its headline: **Steel Alternative.**\n\nI lead growth at Steel. We build browser infrastructure for AI agents. We spend nothing on Google Ads, so the top of our own results page belongs to whoever pays for it.\n\nFirst I was annoyed. Then I got curious and pulled our data to see what this costs us.\n\n## [#](#the-rules-of-the-game) The rules of the game\n\nGoogle lets anyone bid on any brand name as a keyword. It is allowed, and it is normal.\n\nI do think it is a weak bet for developer tools in 2026. The money has a better place to go.\n\n## [#](#quick-napkin-math) Quick napkin math\n\nSay a competitor’s name gets 1,000 searches a month. Your ad shows on most of them.\n\nPeople who search a brand name want that brand. They look for the login page, the docs, the pricing. Few of them click a rival’s ad. Say 2 in 100. That gives you about 15 clicks.\n\nGoogle grades each ad on how well it matches the search. An ad for a different product on a brand search gets a low grade, and a low grade raises your price per click. Say $10.\n\nSo 15 clicks cost $150. If 1 in 20 of those people signs up, you paid about $200 for one signup. That signup may never pay you back.\n\nThe brand owner plays a different game. Their ad matches the search best, so they defend their name for less. They pay less and take the top spot. You pay more and take second place.\n\nThen the owner gets annoyed and starts to bid on your name. Now you both pay, nobody moves, and Google keeps the difference.\n\n## [#](#everyone-in-the-chain-gets-paid) Everyone in the chain gets paid\n\nIt is easy to start. A startup with a fresh round has money and wants growth it can show on a chart. Many hire an agency to run paid search. Agencies are often paid as a share of ad spend, so a bigger budget is good for the agency, even when it is not good for you. Competitor keywords are an easy thing to add to the plan.\n\nGoogle helps too. New advertisers get setup help, account managers, and often some free ad credit to get going. All of that help comes from the company that earns money on every click.\n\nNone of these people are acting in bad faith. But nobody in that chain has a reason to ask the hard question. That question belongs to the founders and the growth lead. Ask it before the first campaign goes live, not after the first invoice.\n\n## [#](#what-happened-when-we-did-nothing) What happened when we did nothing\n\nWe did not defend our name and we did not hit back. We kept building and kept writing.\n\nOver the past year, clicks from searches for our name grew about six times. Since December, our click rate on those searches stayed between 33% and 42% every month.\n\nI can’t prove the ads cost us nothing, because I don’t know the day they started.\n\nThen I looked at where our visitors come from.\n\n## [#](#the-visitor-our-analytics-cant-see) The visitor our analytics can’t see\n\nOur analytics tool says AI assistants send about 3% of our sessions. It says ChatGPT sends almost all of them. Claude barely appears.\n\nI didn’t believe it. Our users are developers, and many of them live in Claude Code all day.\n\nSo I exported raw server logs for two separate 24-hour windows, with the user agent for every request, and I counted.\n\n- For every 5 to 10 homepage loads in a normal browser, AI agents made 1 more request on behalf of a user.\n- Claude Code fetched our site more than 10 times as often as Googlebot did.\n- Claude made 4 to 5 times more requests than ChatGPT.\n\nClaude Code names itself in every request. Codex does not. By default it reads from OpenAI’s cached search index, so it often never reaches our server. When it does fetch a page, the request looks like plain curl or a generic OpenAI request. The real split between tools is less one-sided than these numbers.\n\nAnalytics told the opposite story. It said ChatGPT sent about 100 times more visits than Claude.\n\nBoth are true.\n\nA ChatGPT user clicks a link. The page loads in a browser, the browser runs our analytics script, and we see a visit.\n\nA coding agent does not click. It fetches the page, reads the text, and takes an answer back to the developer. It does not run JavaScript, so our analytics script never loads. We see nothing.\n\nA growing part of how developers judge our product happens where our dashboard can’t look, and where no ad can follow.\n\nI checked whether this was one script on a loop. It was not. The requests arrived in every hour of the day, from many versions of Claude Code, through servers from San Francisco to Mumbai.\n\nAfter the homepage, the page agents read most was pricing.\n\nThe exports cover two days and only our main site. Agents also read our docs, which run on a separate site, so the real number is probably higher. One agent task can also fetch a page more than once, so a fetch is not a person.\n\n## [#](#the-people-who-click-convert-better) The people who click convert better\n\nPeople who come to us from an AI assistant convert at about twice the rate of people from Google search. Over six months, that channel also grew more than twice as fast as Google search.\n\nIt is still smaller than Google. It is also our fastest growing source, and it brings our best visitors.\n\nWhen a developer clicks a link from an assistant, the agent already did the research. The click comes at the end of the decision.\n\nThe ad slot above our name is still for sale. It is worth a little less every month.\n\n## [#](#what-we-should-do-instead) What we should do instead\n\nThese are the rules I follow now.\n\n### 1. Earn the comparison instead of buying the name\n\nA search for “[Competitor]” comes from someone who already chose. A search for “[Competitor] alternative” comes from someone who is still choosing. Write an honest comparison page for the second person, and say where the rival wins. Developers trust a page that admits a weakness. Agents quote it.\n\n### 2. If you must bid, bid on intent\n\nBid on “alternative,” “vs,” and “pricing.” Never bid on the bare name. Send the click to a real comparison page, not to your homepage.\n\n### 3. Defend your name with the product\n\nIf people search for you and click you, the ads above you matter less than you fear.\n\n### 4. Write for the reader who never clicks\n\nYour next user may never see your homepage. Their agent will read your docs, your error messages, and your llms.txt. Keep them complete and current, and put working code in them. An agent will not be impressed by your hero animation.\n\n### 5. Measure what your dashboard misses\n\nOpen your server logs and count the user agents. You will probably find a channel you did not know you had.\n\n### 6. Spend runway where the decision happens\n\nEvery dollar in a brand bidding war is a dollar you do not spend on docs, examples, or product. Investors gave you that money to build an advantage. A tie at a higher price is not one.\n\n## [#](#back-to-the-screenshot) Back to the screenshot\n\nThree companies pay to stand above our name on Google. I see it in other countries too.\n\nMore of our users now ask an agent first. The agent never sees that results page.\n\nI would rather spend our money on the page the agent reads.\n\nGet more like this in your inbox. Unsubscribe anytime.", "url": "https://wpnews.pro/news/bidding-on-competitor-names-when-buyers-ask-agents", "canonical_source": "https://www.nibzard.com/brand-bids", "published_at": "2026-10-08 09:36:12+00:00", "updated_at": "2026-10-08 09:49:30.400213+00:00", "lang": "en", "topics": ["ai-agents", "ai-crawlers", "ai-search"], "entities": ["Steel", "Google", "Google Ads", "ChatGPT", "Claude", "Claude Code", "OpenAI", "Googlebot"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/bidding-on-competitor-names-when-buyers-ask-agents", "markdown": "https://wpnews.pro/news/bidding-on-competitor-names-when-buyers-ask-agents.md", "text": "https://wpnews.pro/news/bidding-on-competitor-names-when-buyers-ask-agents.txt", "jsonld": "https://wpnews.pro/news/bidding-on-competitor-names-when-buyers-ask-agents.jsonld"}}