A biographical page can become a stronger source than the current offer, especially when it keeps a place, a verb and an old mission in the same sentence.
Outside the Part-Dieu metro, a founder showed me a ChatGPT answer on a phone screen, with the late-afternoon sun making it hard to read. The answer named his engineering office correctly. It placed it correctly in Lyon. Then it added a phrase that made him look up: “local integrator that grew out of a workshop near Part-Dieu.” Integrator had not been the right word for a long time.
The case is composite, but the movement is familiar. The office had rewritten its offer pages after tightening its focus around automation design, supervision and technical scoping for industrial workshops. The offer page was clean. A little dry, perhaps, like many pages written after several meetings. Yet the AI answer pulled the company back toward an older origin story, tucked inside the About page. There, one sentence described the beginnings near Part-Dieu, with general integration work. It had survived the redesign.
A secondary page is not a corridor
The About page is often treated as a corridor between the home page and the contact form. It reassures, tells two or three origins, gives the company a more human tone. During redesigns, it rarely comes first. Offer pages receive the attention: category keywords, service titles, mission examples, contact paths. The biographical story sometimes remains intact because it seems less dangerous.
For a model, that hierarchy is not so obvious. An updated service page can contain short blocks, cautious verbs, phrasing that is hard to reuse. It says something right, but in small pieces. An About page, by contrast, often gives a complete sentence: a name, a place, an origin, an initial mission. That sentence has the shape of proof, even if it describes a past state.
In the case of the Lyon office, the offer page explained the current work. The About page gave a more compact image. Two founders, a first shared space, client workshops, integration experience. A human reader understands the chronology. The reader sees that the company has changed focus. The model, however, may keep the piece that travels most easily.
The machine does not know that a beautiful sentence has become decorative.
The story that sticks too well
The problem was not only Part-Dieu. In Lyon, a district immediately adds texture: offices, movement, the station, quick meetings between two trains. The model could use this to produce an answer that seemed informed. The local marker gave texture to the recommendation, like an old sign that remains legible after the façade has changed.
The biographical sentence had another advantage: it tied the place to the line of work. “Born near Part-Dieu around industrial integration missions” forms a solid block. Next to it, “design of supervision systems and scoping of automation architectures” asks for more effort. The second formulation describes the present better. The first travels better in a generated answer.
This is where proof fog appears. The model more or less guesses the company’s value, but it lacks a current sentence that can carry the recommendation without patchwork. It therefore takes an available piece of proof, even if that proof has aged. The answer holds together, but it tilts the company toward a neighboring trade.
The most awkward verb was simple. The site now said “design.” The answer said “install.” In a commercial conversation, that slide changes everything. Design implies upstream work, architectural decisions, responsibility for constraints. Install pulls the company back toward execution. The client who reads the answer may not notice the gap. The founder feels it immediately.
When place makes a category
An About page becomes a guiding source when it connects the company’s name, place and old line of work more clearly than the offer pages do. This stability helps machine readability, but it can freeze an outdated version of the activity.
Lyon’s districts are not neutral in this mechanism. Part-Dieu calls up images of offices and organization. Gerland more readily brings industry, laboratories and technical buildings back into view. Presqu’île gives a more institutional or commercial tone. Vaise can suggest studios, digital services, hybrid teams. These are neighborhoods of words, not facts about a company. But when the source is thin, the neighborhood starts to do work.
In many entity readings, I therefore look at the sentences where district and line of work touch. An old address in isolation is already a signal. An old address joined to a work verb becomes stronger. “Our first projects were born near Part-Dieu” does not produce the same map as “industrial office near Part-Dieu.” The first sentence tells an origin. The second still locates the company in the present.
One might want to remove every mention of the past. That would be a loss. The origin of a technical office sometimes explains its way of working. Former integration experience can give depth to a current design practice. The fragile point is grammatical: the past must remain in the past, and the present must receive a sentence as solid as the old one.
Rereading About as a technical source
I reread an About page through two lenses. The first remains human: does the story create trust without inventing too smooth a hero? The second is more mechanical: which fragments could the model reuse without checking the rest? This second lens may seem cold. It protects against many misunderstandings.
Verbs are often the first clues. “We integrate,” “we deploy,” “we support,” “we design,” “we scope” do not place the company in the same category. On a biographical page, an old verb can remain attached to the company name long after the redesign. It only has to be clearer than the current formulations.
Then I look at how missions are named. A very concrete old intervention can become stronger than several current paragraphs if it gives the model a simple object. In this composite case, a historic mission around equipment integration kept returning as the dominant proof. The current site talked about technical scoping, but without a compact enough example. I would have preferred a steadier sentence: “The office mainly intervenes before deployment, when the workshop has to define the automation architecture, supervision constraints and integration boundaries.”
It is not brilliant. It holds.
The right move is rarely to smooth the whole page. I prefer to add a visible seam. Say where the company comes from, then spell out what that origin no longer means. The About page keeps its narrative grain, but it stops pushing the model toward the wrong category.
Test reused fragments before rewriting
Before touching the text, I observe several answers. A single answer can come from a prompt that was too vague or from a passing association. I note the questions a client would really ask: “automation engineering office Lyon,” “workshop supervision specialist Lyon,” “provider for industrial automation scoping.” Then I check whether the same biographical fragment returns.
When the About page moves the company, the pattern repeats with small variations. Once, the model talks about integration. Another time, it insists on Part-Dieu. A third time, it names design correctly, then adds a sentence about installation. Taken separately, each gap seems tolerable. Together, they draw the wrong map.
The correction should remain modest. The point is not to write in order to hypnotize the machine. It is to reduce the chances that the model takes the wrong bridge. One sentence about the origin, one sentence about the current line of work, one sentence about the boundary with neighboring trades. The rest can breathe. A fully locked text quickly becomes unreadable for clients, and clients remain the first readers.
A company’s machine readability is its capacity to be recognized, located and distinguished in model-generated answers. The About page often plays a larger part in that than one might imagine, because it builds the story that offer pages sometimes leave in pieces.
Note de quai. I keep three traces here: a biographical sentence that was too stable, a verb that pulled the company back toward integration, and an About page more memorable than the current offer. The model’s answer did not come from nowhere; it followed an old sign that was still legible. To make the quay less foggy, I would begin by stitching origin, district and present line of work back together.