Credible Roots / SEO for Founders
Technical SEO, data, and the writing on top of it
SEO for founders, with the working shown
Most SEO retainers report on themselves. Ours are built so a client, a competitor or a journalist can open the site and check the numbers: named data sources, stated assumptions, and a public log of every time we got something wrong.
Client work, published with permission, and all four are checkable on the live site right now. The full case study, including what that correction cost.
SEO for a founder is not a content quota. It is three jobs done together: making a site that search engines and answer engines can crawl, render and resolve to a real entity; publishing material that answers a specific question better than the pages currently ranking for it; and building the kind of evidence that makes both durable, which usually means original data rather than another opinion post. The part almost nobody does is the third. On heatpumpmath.com, a client site we run the SEO for, the differentiator is not the writing, it is that every price comes from the U.S. Energy Information Administration API on a fixed schedule, every constant is printed on the methodology page, and every error is logged publicly with the old and new numbers side by side. When we found that summer gas prices had inflated our own results, the correction cut the number of states where heat pumps won on running cost from nineteen to four, and we published that. Original, checkable data is what search engines reward and what answer engines quote, and it is the one thing a competitor cannot copy in an afternoon.
- Crawlable, renderable, resolvable. Three technical conditions before writing matters at all.
- Original data beats another opinion post, and it is the part that compounds.
- Publish the method and the corrections. It is a ranking asset, not an admission.
- SEO and AEO are related jobs, not the same job. Both are in scope; they are measured differently.
On this page
What most founder SEO is missing
Nearly every stalled SEO engagement we are asked to look at has the same shape. There is a keyword list. There is a content calendar. There are monthly posts. What is missing is anything a reader could not have got from the other nine results, and anything a machine could resolve to a specific, trustworthy entity.
That produces a particular kind of failure. The pages get indexed. They do not get chosen. A search engine ranking ten near-identical explanations of the same topic has no reason to prefer the eleventh, and an answer engine composing a reply has no reason to cite a page whose claims appear verbatim on four other domains. Volume was never the constraint. Being worth choosing was.
The second failure is quieter and more common than it should be. A site that renders only in JavaScript, or blocks the crawlers that feed answer engines, or serves every page under a title the same as forty others, is not competing badly. It is not competing. These are host settings and template defaults that nobody chose deliberately, and they are usually fixable in a day.
What we actually build
Five things, in this order, because each one is wasted without the one before it.
The technical floor
Crawl and render checks, an XML sitemap that matches what actually exists, canonical URLs, correct status codes on retired pages, and a check that the site reads with JavaScript disabled. Then the crawler question, which is now two questions: can Googlebot fetch this, and can GPTBot, ClaudeBot, PerplexityBot and Google-Extended fetch it too. Which of those you should allow is a decision, not a default.
Entity resolution
Structured data that describes a real person and a real organization rather than a generic web page, a canonical identity page, consistent details wherever else you appear, and a Wikidata record where you qualify. Search engines and answer engines both work in entities now. A site that cannot be resolved to one competes only on strings.
Search intent, read backwards
For every target query we look at what is currently ranking and what format it takes, because that is Google telling you what it believes the query wants. If the first page is calculators and you publish an essay, the essay was never in the running. This decides page type before a word is written.
Original data, where there is data to be had
The most defensible page in any niche is the one holding a number nobody else has. Sometimes that means a public API nobody in the sector has bothered to pull. Sometimes it means measuring something yourself. It is more work than an opinion piece and it is the only part competitors cannot replicate the week after they notice you.
Method, assumptions and corrections, in public
Every figure names its source. Every model prints its constants. Every error gets a dated entry with the old and new numbers side by side. This is the ordinary practice of anyone who publishes numbers for a living, it is unusual enough in this market to be a differentiator, and it is exactly the material that answer engines quote.
A worked example: heatpumpmath.com
Named client, published with permission
heatpumpmath.com answers one question: does a heat pump cost less to run than the system already in the house. The obvious way to build that is a calculator on national average prices. We did not build that, because national averages are the reason most of these tools are wrong.
Residential electricity and gas prices come from the U.S. Energy Information Administration's Open Data API v2, pulled by a script that lives in the project, covering 51 states including the District of Columbia. The heating-season window is fixed at December to February for a stated reason: residential gas volume collapses in summer, fixed customer charges then dominate the per-unit price, and using the most recent month would bias every comparison toward heat pumps. Four assumptions are printed with their values. No winner is declared inside a 10% margin. Six states are excluded from gas comparisons where the data does not support one. The computed figures are released under CC BY 4.0.
The part worth reading is the corrections log. On 31 August 2026 we published that the pipeline had been using the most recent monthly figures, which for gas fell in May and June, and that this had inflated the results. Correcting it dropped the number of states where a heat pump wins on running cost under cold conditions from nineteen to four. That correction made the site's own headline finding substantially weaker, and it is on the site with both numbers side by side.
We lead with this rather than a traffic chart because it is the thing an agency cannot fake in a pitch. Anyone can show a graph going up. Very few SEO providers can point at a live methodology page, a live corrections log, and a data pipeline you can go and check.
How to tell real SEO work from a retainer that reports on itself
Four questions worth asking any provider, including us. They are cheap to ask and the answers separate the field quickly.
- Show me a page you would defend under scrutiny. Not the best-performing page: the one whose claims you would stand behind if the subject's competitor read it closely.
- What have you been wrong about, and where is that written down? Anyone who has published for a year has been wrong. A provider with no corrections either has not published enough or does not correct.
- What in this plan could a competitor copy in a week? Formatting, headings and internal links are table stakes. If the whole plan is copyable, it will be copied.
- Which numbers here are yours, and which came from someone else's blog? A statistic passed between marketing posts until nobody can find the original is not evidence. The 7-11-4 rule is the clearest example in this industry.
SEO and AEO are not the same job
They share a technical floor and diverge after it. Search engines rank pages against a query. Answer engines retrieve passages at the moment of asking, compose a reply and cite a few sources. Work that helps one usually helps the other, but they fail differently and they are measured differently, so it is worth being explicit about which you are buying.
| Search (SEO) | Answer engines (AEO) | |
|---|---|---|
| The unit | A page, ranked against a query | A passage, retrieved and quoted |
| What wins | Best match for the intent behind the query | A liftable, specific, corroborated claim |
| Access controlled by | robots.txt and rendering | Separate AI crawler directives, set independently |
| Corroboration | Helps | Close to decisive |
| Measurement | Impressions, position, clicks | Directional: who gets named, and which sources are cited |
Both are in scope here. The AEO half is documented in full on the AI search visibility page, including what we will not promise about it, and the mechanics are written up in how to get cited by ChatGPT, Perplexity and Gemini.
Who this is for
- Founders and operators with a real subject. You have run the thing you would be writing about, and you can tell a reader something that is not on the first page of results already.
- Businesses sitting on data nobody has published. Usage figures, pricing patterns, failure rates, a public API in your sector that nobody has bothered to pull. This is the strongest starting position there is.
- Sites where the technical floor is the bottleneck. Good material that is not being crawled, rendered or resolved. Often the fastest improvement available.
- Anyone who will be checked. If your buyers are technical, or your claims are the product, the method page and the corrections log stop being overhead and start being the reason people believe you.
Who this is not for
Some calls end with us saying this is the wrong piece of work. Cheaper for everyone if that happens before the invoice rather than after it.
- A guaranteed position. Nobody controls the ranking. A provider promising one is either misinformed or counting on you not checking.
- Volume for its own sake. If the brief is a fixed number of posts a month and the topic is whatever ranks, another provider will do that more cheaply and it will work about as well as it usually does.
- A subject you will not stand behind. This method puts your name, your assumptions and your errors in public. That is the mechanism, not a side effect.
- Results this quarter. Technical fixes can move quickly. Earning the position for a competitive query does not, and anyone quoting you a date for it is guessing.
Questions people actually ask
How is this different from an SEO agency retainer?
Mostly in what gets produced. A typical retainer buys a fixed number of posts against a keyword list. This buys a technical floor, an entity that search and answer engines can resolve, and pages built on something you own — data, measurements, a method — that a competitor cannot reproduce by reading your site. The published methodology and corrections log are part of the deliverable rather than internal documents.
Do I need original data to do this?
No, but it is the strongest version. Where there is no data to gather, the differentiator becomes specificity and first-hand experience: the details someone who has actually done the work knows and a generalist writer does not. What does not work is a well-formatted restatement of what already ranks.
Why publish a corrections log? Does that not undermine the site?
It is the opposite in practice. A correction published with the old and new numbers side by side tells a reader the figures are maintained by someone who checks them. Quietly editing an error out tells them nothing, and if anyone noticed the original, it tells them something worse. Search engines and answer engines both prefer sources that look maintained.
How long before this shows up?
Technical problems, once fixed, resolve on the crawler's schedule, which is usually days to weeks. A page earning a competitive position is a matter of months and depends on what it is competing with. We will give you a read on which of the two your situation mostly is on the first call, because they are different pieces of work with different timelines.
Can you work on a site you did not build?
Yes, and that is the usual case. The first pass is a crawl, render and entity audit of what exists, which tells us whether the constraint is technical, editorial or competitive before anything gets rewritten.
Is this included in the monthly package?
Yes. SEO and AI search visibility are part of the same engagement as distribution and the record work, because they run on the same material. What that costs is set out in full.
Read the method before you buy it
- The heatpumpmath.com case study, in full
- AI search visibility: the AEO half of this work
- How to get cited by ChatGPT, Perplexity and Gemini
- Which AI crawlers should you allow
- The site this runs on, built as an entity home
- A worked example of a statistic nobody can source
- Answer engine optimization for companies
- How AI search visibility is actually measured
- The AEO agency field, assessed
- What it costs, published
Sources
- U.S. Energy Information Administration. Open Data API v2. The source behind the heatpumpmath.com pricing pipeline.
- heatpumpmath.com. Methodology: constants, assumptions and where the model fails.
- heatpumpmath.com. Corrections log.
- Google. Search Essentials. The technical, spam and content requirements Google publishes for itself.
- Google. Introduction to structured data markup.
- Creative Commons. Attribution 4.0 International (CC BY 4.0). The licence on the computed figures.
How this was made: written from the work we do for clients, with every figure on this page checkable on the client site it describes. Drafted with AI assistance, then reviewed and approved by the named author before publication. Our editorial standards.
Bring us a site and a subject
The first call is an audit, not a pitch. We will tell you whether your constraint is technical, editorial or competitive, and whether there is data in your sector worth being the first to publish.
Book a 30 minute call