Lightening Tech Evidence, not opinion

SEO & Organic Growth

Being named and cited when someone asks an AI instead of searching

A measured baseline of what ChatGPT, Perplexity and Google's AI answers currently say about you, then monthly work on the sources those answers are actually built from.

From
$600–$4,000/mo
Typical time
ongoing
Usually bought by
Forward-looking brands

Two different people ask for this, and one of them is easier to help

The first arrives frightened. Blog traffic has been sliding for a year, someone has told them AI killed search, and they want the bleeding stopped. The second arrives curious. They typed a question from their own industry into ChatGPT, watched a competitor get named, and wanted to know why it was not them. The second version of this conversation goes better, because it starts from a specific gap rather than from a general dread.

The underlying change is real and worth stating plainly. Google began showing written AI answers above results, then a conversational mode, and at the same time people started taking questions straight to ChatGPT, Perplexity and Claude. On questions that a paragraph resolves, clicks have fallen. Different studies report very different figures for how far, they disagree by industry, and we will not hand you one number and call it a fact. On questions where someone intends to buy something, book something or visit somewhere, clicks have held up, because a summary does not fit a boiler or ship an order.

So the aim of this work is not traffic in the old sense. It is being named, and being cited, in the place where the answer is now given. The name for it is new — AEO, GEO, whichever acronym is being used this quarter — but half the work underneath it is ordinary good SEO, and the other half is public presence in the places these systems read from.

How the work runs, month by month

  1. Build the question set and take a baseline. Thirty to fifty questions your customers genuinely ask, in their words rather than in keyword form, run across ChatGPT, Perplexity and Google's AI mode. For every answer we log whether you were named, which sources were cited, and whether the claims made about you were accurate. That log is the measurement instrument for everything after it. Five to seven days in the first month.
  2. Audit the sources the answers are actually built from. Read the citations and you will usually find the same handful of places — a forum thread, an encyclopaedia entry, an old industry directory, somebody's list article about the best providers in your category. If you are absent there, the system has nowhere to learn your name from. This overlaps heavily with digital PR, and it is the slowest and most valuable part. One week to map, then continuous.
  3. Restructure the pages so they can be quoted. A clear question, an answer that completes inside one paragraph, comparison tables with real values, specific numbers rather than adjectives. Machines lift things that are self-contained. A page that buries the answer in paragraph nine of a narrative gets read and not cited. Two to four pages a month, ongoing.
  4. Settle the machine-readable layer and the crawler decision. Structured data and clean HTML so the meaning of a page is available without inference, and an explicit decision in robots.txt about which AI crawlers may read the site. That decision is yours, it has a real cost either way, and it gets documented rather than defaulted. Two to three days, revisited when policies change.
  5. Re-run the question set and report the delta. Same questions, same systems, same recording format, dated. Alongside it, manual incognito checks on the queries where your position is healthy but clicks are not, because that gap is usually an AI answer sitting above you rather than a ranking problem. One day per month.

What sets the fee

The range is $600–$4,000/mo and it is ongoing because these systems change underneath you.

How wide the question set has to be. A single product in one market needs thirty questions. A business with four product lines across three countries needs a set per line per market, and the monthly re-run is the recurring cost, not the setup.

How much source work is included. Fixing the pages we control is the cheap half. Getting your business represented accurately in the third-party places these systems cite is outreach work, it fails more often than it succeeds, and it belongs at the upper end of the range when it is in scope.

Whether the content work is included or handed over. Restructuring existing pages so they answer cleanly is faster than writing new ones. If the month's work includes producing the material as well as specifying it, the retainer reflects that.

Where this goes wrong

Somebody sells a formula. This field is young enough that confident methodology is a warning sign. There is no verified ranking system to optimise against, no report to point at, and no supplier who can promise your name appears in an answer. The honest version of this service is measured, repeated observation plus work on the inputs — and saying which parts are still experimental, llms.txt included.

The crawlers are blocked and the citations are still expected. A robots.txt rule added during a content-scraping panic, then a retainer bought to increase AI visibility. The two cancel out. Worse, the systems still describe the business, using a five-year-old directory entry and a forum thread, and nobody notices until a customer repeats the wrong price back to the sales team.

Informational pages get rewritten because their clicks fell. A page holding position three with almost no clicks looks broken and usually is not. Check the result page by hand and you will often find an AI answer occupying the top of the screen for that query and four neighbouring ones. The ranking is healthy and the click opportunity is gone. Rewriting the page does not bring it back, and deleting it removes the page the AI answer was drawing your brand mention from. The correct move is to keep it and shift effort to queries where a click still converts.

See where the site stands before you brief anyone.

If the site is already live, start from what it scores rather than from a guess. The audit reports what it checked, what it found, and — the part most tools leave out — what it could not check, and why.

See the website audit Talk about this work

Questions people ask before they commit.

Can you guarantee we appear in AI Overviews or in ChatGPT answers?
No, and anyone selling a formula for it is selling something they cannot control. These systems are not ranked lists you can hold a position in, and they change without notice. What we commit to is the measurable part — a fixed set of questions run on a schedule, a record of what each system said and which sources it drew on, and work on the sources and pages that feed those answers. The movement in the answers themselves is a goal, not a promise.
How do you measure something with no ranking report?
With a question set and a log. We build thirty to fifty questions your actual customers ask, run them across ChatGPT, Perplexity and Google's AI mode, and record three things for each answer — whether your brand was named, which sources were cited, and whether what was said about you was correct. Re-running the same set on the same schedule is what turns an opinion into a trend. That third column matters more than people expect, because being described wrongly is worse than being absent.
We block AI crawlers to protect our content. Does that affect this?
Directly, and it is a decision you have to make with the trade-off in front of you rather than by default. A system that cannot read your pages will describe you using whatever else it can find — an old directory listing, a forum thread, a competitor's comparison article. Blocking is a legitimate choice for some businesses, but you cannot block access and expect to be cited accurately, and we would rather have that conversation at the start than after three months of work.
Our blog traffic has fallen. Is this the service that fixes it?
Sometimes, and sometimes the honest answer is that the traffic is not coming back. Clicks have dropped hardest on questions a written answer settles in a paragraph, while searches where someone intends to buy, book or visit have held up. If your fall is concentrated on informational pages holding good positions, the ranking is not broken and the fix is not to rewrite those pages. It is to move effort toward the queries where a click is still worth having, and to make sure the informational pages earn a mention inside the answer instead.
Is llms.txt worth doing?
It costs about half an hour and there is no proven benefit. It was proposed in 2024 as a summary file for AI systems, no major provider has confirmed using it, and Google has publicly expressed scepticism. We will add one because it is cheap and harmless, and we will not present it as a result. Anyone charging meaningfully for it is charging for an experiment.
Does Search Console show us any of this?
Not separately, and this is a limitation worth knowing before you read any report. There is no AI Overview filter — those impressions and clicks are mixed into ordinary web data. Where we need to know whether an AI answer sits above your result, we check the query by hand in a fresh incognito window and date the observation. It is a manual method and it is the only one that tells the truth right now.

Often taken together.

Start with the brief, not the invoice.

Send what you are trying to achieve. You get a written scope, a price and a date — or an honest “this is not the right service for that”.

Talk about AI search / AEO optimisation All SEO & Organic Growth services